diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,27976 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 7.0, + "eval_steps": 500, + "global_step": 12698, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0027563395810363835, + "grad_norm": 12.03199807971177, + "learning_rate": 1.2598425196850394e-07, + "loss": 0.5457, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25345396995544434, + "step": 5, + "valid_targets_mean": 7359.2, + "valid_targets_min": 5595 + }, + { + "epoch": 0.005512679162072767, + "grad_norm": 10.8365471165428, + "learning_rate": 2.8346456692913386e-07, + "loss": 0.5804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1994602382183075, + "step": 10, + "valid_targets_mean": 1287.1, + "valid_targets_min": 420 + }, + { + "epoch": 0.008269018743109152, + "grad_norm": 9.9807762660995, + "learning_rate": 4.409448818897638e-07, + "loss": 0.5587, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2705960273742676, + "step": 15, + "valid_targets_mean": 8512.8, + "valid_targets_min": 6729 + }, + { + "epoch": 0.011025358324145534, + "grad_norm": 9.512689549073038, + "learning_rate": 5.984251968503937e-07, + "loss": 0.5105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2382660061120987, + "step": 20, + "valid_targets_mean": 7552.0, + "valid_targets_min": 6145 + }, + { + "epoch": 0.013781697905181918, + "grad_norm": 7.959840819817274, + "learning_rate": 7.559055118110237e-07, + "loss": 0.5233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2503214478492737, + "step": 25, + "valid_targets_mean": 7301.4, + "valid_targets_min": 4807 + }, + { + "epoch": 0.016538037486218304, + "grad_norm": 6.403061978421631, + "learning_rate": 9.133858267716536e-07, + "loss": 0.5295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28035780787467957, + "step": 30, + "valid_targets_mean": 6809.0, + "valid_targets_min": 5015 + }, + { + "epoch": 0.019294377067254686, + "grad_norm": 4.970118339440747, + "learning_rate": 1.0708661417322836e-06, + "loss": 0.5218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2603369653224945, + "step": 35, + "valid_targets_mean": 7217.0, + "valid_targets_min": 5541 + }, + { + "epoch": 0.022050716648291068, + "grad_norm": 4.290559207246597, + "learning_rate": 1.2283464566929135e-06, + "loss": 0.4927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23547153174877167, + "step": 40, + "valid_targets_mean": 6205.1, + "valid_targets_min": 5580 + }, + { + "epoch": 0.024807056229327454, + "grad_norm": 5.621259235341147, + "learning_rate": 1.3858267716535435e-06, + "loss": 0.5003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11650706082582474, + "step": 45, + "valid_targets_mean": 598.0, + "valid_targets_min": 151 + }, + { + "epoch": 0.027563395810363836, + "grad_norm": 2.597053882413677, + "learning_rate": 1.5433070866141734e-06, + "loss": 0.4538, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22514992952346802, + "step": 50, + "valid_targets_mean": 7270.8, + "valid_targets_min": 6125 + }, + { + "epoch": 0.03031973539140022, + "grad_norm": 2.2144328767407466, + "learning_rate": 1.7007874015748034e-06, + "loss": 0.4519, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21197445690631866, + "step": 55, + "valid_targets_mean": 6108.9, + "valid_targets_min": 4623 + }, + { + "epoch": 0.03307607497243661, + "grad_norm": 1.748563322135922, + "learning_rate": 1.8582677165354333e-06, + "loss": 0.4359, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2161049246788025, + "step": 60, + "valid_targets_mean": 6571.6, + "valid_targets_min": 4850 + }, + { + "epoch": 0.03583241455347299, + "grad_norm": 1.2883846562207764, + "learning_rate": 2.015748031496063e-06, + "loss": 0.4134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21156634390354156, + "step": 65, + "valid_targets_mean": 7984.4, + "valid_targets_min": 6312 + }, + { + "epoch": 0.03858875413450937, + "grad_norm": 1.1191372312735937, + "learning_rate": 2.173228346456693e-06, + "loss": 0.4015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19570806622505188, + "step": 70, + "valid_targets_mean": 6273.1, + "valid_targets_min": 4590 + }, + { + "epoch": 0.041345093715545754, + "grad_norm": 0.9924950730350816, + "learning_rate": 2.330708661417323e-06, + "loss": 0.4051, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20732605457305908, + "step": 75, + "valid_targets_mean": 7860.2, + "valid_targets_min": 5453 + }, + { + "epoch": 0.044101433296582136, + "grad_norm": 0.8480114058846641, + "learning_rate": 2.488188976377953e-06, + "loss": 0.4007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18277965486049652, + "step": 80, + "valid_targets_mean": 6988.6, + "valid_targets_min": 4928 + }, + { + "epoch": 0.046857772877618525, + "grad_norm": 0.7210109308748683, + "learning_rate": 2.645669291338583e-06, + "loss": 0.387, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1823805570602417, + "step": 85, + "valid_targets_mean": 6495.6, + "valid_targets_min": 5367 + }, + { + "epoch": 0.04961411245865491, + "grad_norm": 0.7912676146342451, + "learning_rate": 2.803149606299213e-06, + "loss": 0.3853, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19350218772888184, + "step": 90, + "valid_targets_mean": 4968.1, + "valid_targets_min": 615 + }, + { + "epoch": 0.05237045203969129, + "grad_norm": 0.6476668115771634, + "learning_rate": 2.9606299212598427e-06, + "loss": 0.3967, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19813990592956543, + "step": 95, + "valid_targets_mean": 7189.9, + "valid_targets_min": 5732 + }, + { + "epoch": 0.05512679162072767, + "grad_norm": 0.5762264226259036, + "learning_rate": 3.118110236220473e-06, + "loss": 0.3832, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1988917887210846, + "step": 100, + "valid_targets_mean": 7969.4, + "valid_targets_min": 5870 + }, + { + "epoch": 0.057883131201764054, + "grad_norm": 0.5563933891999993, + "learning_rate": 3.2755905511811026e-06, + "loss": 0.3462, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1556871235370636, + "step": 105, + "valid_targets_mean": 7348.0, + "valid_targets_min": 4983 + }, + { + "epoch": 0.06063947078280044, + "grad_norm": 0.5492487878055661, + "learning_rate": 3.433070866141733e-06, + "loss": 0.3479, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15495307743549347, + "step": 110, + "valid_targets_mean": 6763.5, + "valid_targets_min": 5251 + }, + { + "epoch": 0.06339581036383682, + "grad_norm": 0.6584365379384388, + "learning_rate": 3.5905511811023625e-06, + "loss": 0.3629, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20552292466163635, + "step": 115, + "valid_targets_mean": 6626.4, + "valid_targets_min": 4332 + }, + { + "epoch": 0.06615214994487321, + "grad_norm": 0.4985276315183511, + "learning_rate": 3.748031496062993e-06, + "loss": 0.3495, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.166792631149292, + "step": 120, + "valid_targets_mean": 7151.6, + "valid_targets_min": 4434 + }, + { + "epoch": 0.0689084895259096, + "grad_norm": 0.5347514600849942, + "learning_rate": 3.905511811023622e-06, + "loss": 0.3432, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19392603635787964, + "step": 125, + "valid_targets_mean": 7592.0, + "valid_targets_min": 5705 + }, + { + "epoch": 0.07166482910694598, + "grad_norm": 0.48836979467828007, + "learning_rate": 4.062992125984252e-06, + "loss": 0.3031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14662516117095947, + "step": 130, + "valid_targets_mean": 8727.9, + "valid_targets_min": 5721 + }, + { + "epoch": 0.07442116868798236, + "grad_norm": 0.5207572085784237, + "learning_rate": 4.220472440944882e-06, + "loss": 0.3019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16591835021972656, + "step": 135, + "valid_targets_mean": 6314.9, + "valid_targets_min": 4517 + }, + { + "epoch": 0.07717750826901874, + "grad_norm": 0.4974724138813966, + "learning_rate": 4.377952755905512e-06, + "loss": 0.2898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16314616799354553, + "step": 140, + "valid_targets_mean": 7161.6, + "valid_targets_min": 4531 + }, + { + "epoch": 0.07993384785005513, + "grad_norm": 1.1371696349422944, + "learning_rate": 4.535433070866142e-06, + "loss": 0.3532, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13186468183994293, + "step": 145, + "valid_targets_mean": 1098.8, + "valid_targets_min": 160 + }, + { + "epoch": 0.08269018743109151, + "grad_norm": 0.663986696798913, + "learning_rate": 4.692913385826772e-06, + "loss": 0.3398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1464945524930954, + "step": 150, + "valid_targets_mean": 6988.4, + "valid_targets_min": 5392 + }, + { + "epoch": 0.08544652701212789, + "grad_norm": 0.5475709313298295, + "learning_rate": 4.850393700787402e-06, + "loss": 0.3342, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1705450415611267, + "step": 155, + "valid_targets_mean": 7712.9, + "valid_targets_min": 5297 + }, + { + "epoch": 0.08820286659316427, + "grad_norm": 0.4966605266774129, + "learning_rate": 5.007874015748032e-06, + "loss": 0.3406, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1669228971004486, + "step": 160, + "valid_targets_mean": 6396.9, + "valid_targets_min": 5536 + }, + { + "epoch": 0.09095920617420065, + "grad_norm": 0.5186459086053145, + "learning_rate": 5.1653543307086615e-06, + "loss": 0.3324, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1706322431564331, + "step": 165, + "valid_targets_mean": 6143.8, + "valid_targets_min": 3434 + }, + { + "epoch": 0.09371554575523705, + "grad_norm": 0.4494525100421949, + "learning_rate": 5.322834645669291e-06, + "loss": 0.3156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1617979258298874, + "step": 170, + "valid_targets_mean": 7182.1, + "valid_targets_min": 5349 + }, + { + "epoch": 0.09647188533627343, + "grad_norm": 0.4522494748694371, + "learning_rate": 5.480314960629922e-06, + "loss": 0.3196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16258108615875244, + "step": 175, + "valid_targets_mean": 7570.5, + "valid_targets_min": 5209 + }, + { + "epoch": 0.09922822491730982, + "grad_norm": 0.5085204524684658, + "learning_rate": 5.637795275590552e-06, + "loss": 0.3109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17022168636322021, + "step": 180, + "valid_targets_mean": 7069.0, + "valid_targets_min": 5581 + }, + { + "epoch": 0.1019845644983462, + "grad_norm": 0.5209437697173587, + "learning_rate": 5.795275590551181e-06, + "loss": 0.3155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16440576314926147, + "step": 185, + "valid_targets_mean": 6925.6, + "valid_targets_min": 5504 + }, + { + "epoch": 0.10474090407938258, + "grad_norm": 0.47995002900832023, + "learning_rate": 5.952755905511811e-06, + "loss": 0.3082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1566929817199707, + "step": 190, + "valid_targets_mean": 6716.9, + "valid_targets_min": 5332 + }, + { + "epoch": 0.10749724366041896, + "grad_norm": 0.5248122553166824, + "learning_rate": 6.110236220472442e-06, + "loss": 0.312, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1520642191171646, + "step": 195, + "valid_targets_mean": 6241.8, + "valid_targets_min": 5406 + }, + { + "epoch": 0.11025358324145534, + "grad_norm": 0.47394348554230464, + "learning_rate": 6.2677165354330715e-06, + "loss": 0.3119, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15291783213615417, + "step": 200, + "valid_targets_mean": 6387.8, + "valid_targets_min": 4847 + }, + { + "epoch": 0.11300992282249173, + "grad_norm": 0.4918866725947959, + "learning_rate": 6.425196850393701e-06, + "loss": 0.3081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15237608551979065, + "step": 205, + "valid_targets_mean": 6457.6, + "valid_targets_min": 4288 + }, + { + "epoch": 0.11576626240352811, + "grad_norm": 0.5092556466274586, + "learning_rate": 6.5826771653543306e-06, + "loss": 0.3075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17032074928283691, + "step": 210, + "valid_targets_mean": 6793.1, + "valid_targets_min": 5173 + }, + { + "epoch": 0.1185226019845645, + "grad_norm": 0.4566280978554122, + "learning_rate": 6.740157480314962e-06, + "loss": 0.3002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13838636875152588, + "step": 215, + "valid_targets_mean": 6697.4, + "valid_targets_min": 5168 + }, + { + "epoch": 0.12127894156560089, + "grad_norm": 0.4729349800083246, + "learning_rate": 6.897637795275591e-06, + "loss": 0.2971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1457296758890152, + "step": 220, + "valid_targets_mean": 6062.0, + "valid_targets_min": 5270 + }, + { + "epoch": 0.12403528114663727, + "grad_norm": 0.48488877298525956, + "learning_rate": 7.055118110236221e-06, + "loss": 0.2997, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15444348752498627, + "step": 225, + "valid_targets_mean": 7110.1, + "valid_targets_min": 5253 + }, + { + "epoch": 0.12679162072767364, + "grad_norm": 0.5077407739257016, + "learning_rate": 7.21259842519685e-06, + "loss": 0.2895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13668900728225708, + "step": 230, + "valid_targets_mean": 7395.9, + "valid_targets_min": 5248 + }, + { + "epoch": 0.12954796030871002, + "grad_norm": 0.6537668754430908, + "learning_rate": 7.3700787401574816e-06, + "loss": 0.2965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1796347200870514, + "step": 235, + "valid_targets_mean": 7033.5, + "valid_targets_min": 5354 + }, + { + "epoch": 0.13230429988974643, + "grad_norm": 0.47144612143290426, + "learning_rate": 7.527559055118111e-06, + "loss": 0.2966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15602543950080872, + "step": 240, + "valid_targets_mean": 6798.8, + "valid_targets_min": 4532 + }, + { + "epoch": 0.1350606394707828, + "grad_norm": 0.919027652758804, + "learning_rate": 7.68503937007874e-06, + "loss": 0.2748, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09504696726799011, + "step": 245, + "valid_targets_mean": 2045.0, + "valid_targets_min": 164 + }, + { + "epoch": 0.1378169790518192, + "grad_norm": 0.45302252768642653, + "learning_rate": 7.84251968503937e-06, + "loss": 0.286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14491286873817444, + "step": 250, + "valid_targets_mean": 6723.1, + "valid_targets_min": 5164 + }, + { + "epoch": 0.14057331863285558, + "grad_norm": 0.4814458529170389, + "learning_rate": 8.000000000000001e-06, + "loss": 0.2859, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14560231566429138, + "step": 255, + "valid_targets_mean": 5954.1, + "valid_targets_min": 4996 + }, + { + "epoch": 0.14332965821389196, + "grad_norm": 0.5398081158021217, + "learning_rate": 8.157480314960631e-06, + "loss": 0.2912, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13352955877780914, + "step": 260, + "valid_targets_mean": 5932.0, + "valid_targets_min": 5205 + }, + { + "epoch": 0.14608599779492834, + "grad_norm": 0.48936367805691844, + "learning_rate": 8.31496062992126e-06, + "loss": 0.2851, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1488025039434433, + "step": 265, + "valid_targets_mean": 6288.4, + "valid_targets_min": 5380 + }, + { + "epoch": 0.14884233737596472, + "grad_norm": 0.4510776733853006, + "learning_rate": 8.47244094488189e-06, + "loss": 0.2702, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11579629778862, + "step": 270, + "valid_targets_mean": 7095.9, + "valid_targets_min": 5048 + }, + { + "epoch": 0.1515986769570011, + "grad_norm": 0.5122626144884862, + "learning_rate": 8.629921259842521e-06, + "loss": 0.2857, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14105123281478882, + "step": 275, + "valid_targets_mean": 6093.5, + "valid_targets_min": 4489 + }, + { + "epoch": 0.1543550165380375, + "grad_norm": 2.730769352259255, + "learning_rate": 8.78740157480315e-06, + "loss": 0.3304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.234264075756073, + "step": 280, + "valid_targets_mean": 1659.9, + "valid_targets_min": 857 + }, + { + "epoch": 0.15711135611907387, + "grad_norm": 1.4424943519130584, + "learning_rate": 8.94488188976378e-06, + "loss": 0.3819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18700096011161804, + "step": 285, + "valid_targets_mean": 1260.4, + "valid_targets_min": 686 + }, + { + "epoch": 0.15986769570011025, + "grad_norm": 1.2666929982937107, + "learning_rate": 9.10236220472441e-06, + "loss": 0.3535, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12993380427360535, + "step": 290, + "valid_targets_mean": 1226.2, + "valid_targets_min": 676 + }, + { + "epoch": 0.16262403528114663, + "grad_norm": 1.2398113220827929, + "learning_rate": 9.259842519685041e-06, + "loss": 0.3333, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14230385422706604, + "step": 295, + "valid_targets_mean": 1129.9, + "valid_targets_min": 661 + }, + { + "epoch": 0.16538037486218302, + "grad_norm": 1.091300115223105, + "learning_rate": 9.41732283464567e-06, + "loss": 0.3258, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15736964344978333, + "step": 300, + "valid_targets_mean": 1631.2, + "valid_targets_min": 520 + }, + { + "epoch": 0.1681367144432194, + "grad_norm": 1.1347979672625979, + "learning_rate": 9.5748031496063e-06, + "loss": 0.3114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1491531878709793, + "step": 305, + "valid_targets_mean": 1091.9, + "valid_targets_min": 472 + }, + { + "epoch": 0.17089305402425578, + "grad_norm": 1.0280814738911594, + "learning_rate": 9.73228346456693e-06, + "loss": 0.3202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1237805187702179, + "step": 310, + "valid_targets_mean": 1244.9, + "valid_targets_min": 497 + }, + { + "epoch": 0.17364939360529216, + "grad_norm": 1.098677879261211, + "learning_rate": 9.88976377952756e-06, + "loss": 0.292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14757400751113892, + "step": 315, + "valid_targets_mean": 1297.4, + "valid_targets_min": 453 + }, + { + "epoch": 0.17640573318632854, + "grad_norm": 1.3269439839421167, + "learning_rate": 1.0047244094488188e-05, + "loss": 0.3184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13808107376098633, + "step": 320, + "valid_targets_mean": 1147.8, + "valid_targets_min": 501 + }, + { + "epoch": 0.17916207276736493, + "grad_norm": 1.0686383329170992, + "learning_rate": 1.0204724409448821e-05, + "loss": 0.294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09450952708721161, + "step": 325, + "valid_targets_mean": 1010.0, + "valid_targets_min": 580 + }, + { + "epoch": 0.1819184123484013, + "grad_norm": 2.5058689007729074, + "learning_rate": 1.0362204724409451e-05, + "loss": 0.2925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19288481771945953, + "step": 330, + "valid_targets_mean": 1627.9, + "valid_targets_min": 698 + }, + { + "epoch": 0.18467475192943772, + "grad_norm": 1.0516739540628306, + "learning_rate": 1.051968503937008e-05, + "loss": 0.2934, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11996397376060486, + "step": 335, + "valid_targets_mean": 1142.4, + "valid_targets_min": 636 + }, + { + "epoch": 0.1874310915104741, + "grad_norm": 1.182224037115777, + "learning_rate": 1.067716535433071e-05, + "loss": 0.2882, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13485287129878998, + "step": 340, + "valid_targets_mean": 1073.8, + "valid_targets_min": 568 + }, + { + "epoch": 0.19018743109151048, + "grad_norm": 0.9817198216542092, + "learning_rate": 1.083464566929134e-05, + "loss": 0.2659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13364845514297485, + "step": 345, + "valid_targets_mean": 1611.5, + "valid_targets_min": 982 + }, + { + "epoch": 0.19294377067254687, + "grad_norm": 1.0578179933528884, + "learning_rate": 1.0992125984251969e-05, + "loss": 0.2666, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1154220700263977, + "step": 350, + "valid_targets_mean": 1190.0, + "valid_targets_min": 921 + }, + { + "epoch": 0.19570011025358325, + "grad_norm": 1.0354377407177526, + "learning_rate": 1.1149606299212599e-05, + "loss": 0.2751, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14310000836849213, + "step": 355, + "valid_targets_mean": 1367.0, + "valid_targets_min": 559 + }, + { + "epoch": 0.19845644983461963, + "grad_norm": 0.993624991070164, + "learning_rate": 1.1307086614173228e-05, + "loss": 0.2744, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18068936467170715, + "step": 360, + "valid_targets_mean": 1773.5, + "valid_targets_min": 609 + }, + { + "epoch": 0.201212789415656, + "grad_norm": 1.133015530814811, + "learning_rate": 1.1464566929133861e-05, + "loss": 0.2838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16420716047286987, + "step": 365, + "valid_targets_mean": 1784.5, + "valid_targets_min": 865 + }, + { + "epoch": 0.2039691289966924, + "grad_norm": 1.1381075467310309, + "learning_rate": 1.162204724409449e-05, + "loss": 0.2708, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1279158890247345, + "step": 370, + "valid_targets_mean": 1142.9, + "valid_targets_min": 858 + }, + { + "epoch": 0.20672546857772878, + "grad_norm": 1.063761556416521, + "learning_rate": 1.177952755905512e-05, + "loss": 0.2778, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15421351790428162, + "step": 375, + "valid_targets_mean": 1556.0, + "valid_targets_min": 461 + }, + { + "epoch": 0.20948180815876516, + "grad_norm": 1.1666487461437867, + "learning_rate": 1.193700787401575e-05, + "loss": 0.274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11098136007785797, + "step": 380, + "valid_targets_mean": 882.1, + "valid_targets_min": 476 + }, + { + "epoch": 0.21223814773980154, + "grad_norm": 1.2197286387894173, + "learning_rate": 1.2094488188976379e-05, + "loss": 0.2821, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13837429881095886, + "step": 385, + "valid_targets_mean": 1290.1, + "valid_targets_min": 469 + }, + { + "epoch": 0.21499448732083792, + "grad_norm": 1.1355737051924608, + "learning_rate": 1.2251968503937009e-05, + "loss": 0.2797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09326247870922089, + "step": 390, + "valid_targets_mean": 1085.0, + "valid_targets_min": 644 + }, + { + "epoch": 0.2177508269018743, + "grad_norm": 1.4585964100100204, + "learning_rate": 1.2409448818897638e-05, + "loss": 0.2533, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17078709602355957, + "step": 395, + "valid_targets_mean": 1322.4, + "valid_targets_min": 673 + }, + { + "epoch": 0.2205071664829107, + "grad_norm": 1.0789587675176802, + "learning_rate": 1.2566929133858268e-05, + "loss": 0.2713, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12583965063095093, + "step": 400, + "valid_targets_mean": 1386.5, + "valid_targets_min": 574 + }, + { + "epoch": 0.22326350606394707, + "grad_norm": 1.0135939374793805, + "learning_rate": 1.2724409448818899e-05, + "loss": 0.2678, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14031243324279785, + "step": 405, + "valid_targets_mean": 1325.2, + "valid_targets_min": 513 + }, + { + "epoch": 0.22601984564498345, + "grad_norm": 1.10790017884695, + "learning_rate": 1.288188976377953e-05, + "loss": 0.2668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0964772030711174, + "step": 410, + "valid_targets_mean": 1024.5, + "valid_targets_min": 670 + }, + { + "epoch": 0.22877618522601983, + "grad_norm": 1.1232137615686137, + "learning_rate": 1.303937007874016e-05, + "loss": 0.2672, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13388940691947937, + "step": 415, + "valid_targets_mean": 1211.0, + "valid_targets_min": 583 + }, + { + "epoch": 0.23153252480705622, + "grad_norm": 1.1006858734350895, + "learning_rate": 1.3196850393700789e-05, + "loss": 0.2481, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10147290676832199, + "step": 420, + "valid_targets_mean": 1101.4, + "valid_targets_min": 503 + }, + { + "epoch": 0.2342888643880926, + "grad_norm": 1.301922760962984, + "learning_rate": 1.3354330708661419e-05, + "loss": 0.2591, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09904228150844574, + "step": 425, + "valid_targets_mean": 995.1, + "valid_targets_min": 642 + }, + { + "epoch": 0.237045203969129, + "grad_norm": 1.0854691830530072, + "learning_rate": 1.3511811023622048e-05, + "loss": 0.2624, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11032722145318985, + "step": 430, + "valid_targets_mean": 1149.6, + "valid_targets_min": 530 + }, + { + "epoch": 0.2398015435501654, + "grad_norm": 1.0716451640647942, + "learning_rate": 1.3669291338582678e-05, + "loss": 0.2628, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11942590028047562, + "step": 435, + "valid_targets_mean": 1142.5, + "valid_targets_min": 805 + }, + { + "epoch": 0.24255788313120177, + "grad_norm": 1.0262801903372485, + "learning_rate": 1.3826771653543307e-05, + "loss": 0.2643, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12179580330848694, + "step": 440, + "valid_targets_mean": 1230.5, + "valid_targets_min": 498 + }, + { + "epoch": 0.24531422271223816, + "grad_norm": 1.3355224484466277, + "learning_rate": 1.3984251968503938e-05, + "loss": 0.2505, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09617472440004349, + "step": 445, + "valid_targets_mean": 1020.4, + "valid_targets_min": 658 + }, + { + "epoch": 0.24807056229327454, + "grad_norm": 1.0418667407730249, + "learning_rate": 1.4141732283464568e-05, + "loss": 0.2351, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10658672451972961, + "step": 450, + "valid_targets_mean": 1393.4, + "valid_targets_min": 594 + }, + { + "epoch": 0.2508269018743109, + "grad_norm": 1.0186190918836215, + "learning_rate": 1.4299212598425199e-05, + "loss": 0.2613, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14161312580108643, + "step": 455, + "valid_targets_mean": 1613.1, + "valid_targets_min": 674 + }, + { + "epoch": 0.2535832414553473, + "grad_norm": 1.0198559009793984, + "learning_rate": 1.4456692913385829e-05, + "loss": 0.2585, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08782072365283966, + "step": 460, + "valid_targets_mean": 1025.6, + "valid_targets_min": 648 + }, + { + "epoch": 0.25633958103638366, + "grad_norm": 1.0453836775655403, + "learning_rate": 1.4614173228346458e-05, + "loss": 0.2439, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13071300089359283, + "step": 465, + "valid_targets_mean": 1291.6, + "valid_targets_min": 545 + }, + { + "epoch": 0.25909592061742004, + "grad_norm": 1.4827307882748804, + "learning_rate": 1.4771653543307088e-05, + "loss": 0.2458, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12336685508489609, + "step": 470, + "valid_targets_mean": 1272.5, + "valid_targets_min": 736 + }, + { + "epoch": 0.2618522601984565, + "grad_norm": 1.0038709930452714, + "learning_rate": 1.4929133858267717e-05, + "loss": 0.2589, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14301827549934387, + "step": 475, + "valid_targets_mean": 1435.8, + "valid_targets_min": 547 + }, + { + "epoch": 0.26460859977949286, + "grad_norm": 1.1056058255122372, + "learning_rate": 1.5086614173228347e-05, + "loss": 0.2433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10720009356737137, + "step": 480, + "valid_targets_mean": 1166.6, + "valid_targets_min": 737 + }, + { + "epoch": 0.26736493936052924, + "grad_norm": 1.0242750741774154, + "learning_rate": 1.5244094488188978e-05, + "loss": 0.2592, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13384518027305603, + "step": 485, + "valid_targets_mean": 1414.2, + "valid_targets_min": 764 + }, + { + "epoch": 0.2701212789415656, + "grad_norm": 1.0115174147228072, + "learning_rate": 1.5401574803149607e-05, + "loss": 0.2649, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13870371878147125, + "step": 490, + "valid_targets_mean": 1457.8, + "valid_targets_min": 662 + }, + { + "epoch": 0.272877618522602, + "grad_norm": 1.0812192908699654, + "learning_rate": 1.555905511811024e-05, + "loss": 0.2642, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11753414571285248, + "step": 495, + "valid_targets_mean": 1344.4, + "valid_targets_min": 680 + }, + { + "epoch": 0.2756339581036384, + "grad_norm": 1.0812003513311095, + "learning_rate": 1.5716535433070866e-05, + "loss": 0.2457, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11089211702346802, + "step": 500, + "valid_targets_mean": 1200.9, + "valid_targets_min": 641 + }, + { + "epoch": 0.27839029768467477, + "grad_norm": 1.0471042142990983, + "learning_rate": 1.5874015748031498e-05, + "loss": 0.2483, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1205059140920639, + "step": 505, + "valid_targets_mean": 1432.2, + "valid_targets_min": 730 + }, + { + "epoch": 0.28114663726571115, + "grad_norm": 1.0040363937518773, + "learning_rate": 1.6031496062992126e-05, + "loss": 0.2559, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13119642436504364, + "step": 510, + "valid_targets_mean": 1447.6, + "valid_targets_min": 598 + }, + { + "epoch": 0.28390297684674753, + "grad_norm": 1.0737043456043254, + "learning_rate": 1.6188976377952757e-05, + "loss": 0.2413, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1225944459438324, + "step": 515, + "valid_targets_mean": 1212.6, + "valid_targets_min": 583 + }, + { + "epoch": 0.2866593164277839, + "grad_norm": 1.1022856222478425, + "learning_rate": 1.6346456692913385e-05, + "loss": 0.2391, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09562849253416061, + "step": 520, + "valid_targets_mean": 1051.5, + "valid_targets_min": 605 + }, + { + "epoch": 0.2894156560088203, + "grad_norm": 1.0130694551675619, + "learning_rate": 1.650393700787402e-05, + "loss": 0.2446, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14441713690757751, + "step": 525, + "valid_targets_mean": 1602.2, + "valid_targets_min": 613 + }, + { + "epoch": 0.2921719955898567, + "grad_norm": 1.0094303193906469, + "learning_rate": 1.6661417322834647e-05, + "loss": 0.2451, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15729200839996338, + "step": 530, + "valid_targets_mean": 1667.2, + "valid_targets_min": 578 + }, + { + "epoch": 0.29492833517089306, + "grad_norm": 0.9867793871407218, + "learning_rate": 1.6818897637795278e-05, + "loss": 0.2468, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1478235423564911, + "step": 535, + "valid_targets_mean": 1637.0, + "valid_targets_min": 497 + }, + { + "epoch": 0.29768467475192945, + "grad_norm": 0.9497069693079443, + "learning_rate": 1.6976377952755906e-05, + "loss": 0.2406, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14882995188236237, + "step": 540, + "valid_targets_mean": 1513.6, + "valid_targets_min": 603 + }, + { + "epoch": 0.30044101433296583, + "grad_norm": 0.9375744348491698, + "learning_rate": 1.7133858267716537e-05, + "loss": 0.2374, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13180536031723022, + "step": 545, + "valid_targets_mean": 1785.1, + "valid_targets_min": 520 + }, + { + "epoch": 0.3031973539140022, + "grad_norm": 1.1028464926940709, + "learning_rate": 1.7291338582677165e-05, + "loss": 0.2453, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11303818970918655, + "step": 550, + "valid_targets_mean": 1283.2, + "valid_targets_min": 593 + }, + { + "epoch": 0.3059536934950386, + "grad_norm": 0.9800687897793249, + "learning_rate": 1.7448818897637796e-05, + "loss": 0.2426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1307448148727417, + "step": 555, + "valid_targets_mean": 1396.6, + "valid_targets_min": 689 + }, + { + "epoch": 0.308710033076075, + "grad_norm": 1.1586783542022199, + "learning_rate": 1.7606299212598424e-05, + "loss": 0.2409, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13448111712932587, + "step": 560, + "valid_targets_mean": 1385.5, + "valid_targets_min": 577 + }, + { + "epoch": 0.31146637265711136, + "grad_norm": 1.059161978260715, + "learning_rate": 1.776377952755906e-05, + "loss": 0.2281, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10515008121728897, + "step": 565, + "valid_targets_mean": 1074.1, + "valid_targets_min": 511 + }, + { + "epoch": 0.31422271223814774, + "grad_norm": 0.9894226659961185, + "learning_rate": 1.7921259842519687e-05, + "loss": 0.2369, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11276115477085114, + "step": 570, + "valid_targets_mean": 1384.9, + "valid_targets_min": 777 + }, + { + "epoch": 0.3169790518191841, + "grad_norm": 0.9945432011152454, + "learning_rate": 1.8078740157480318e-05, + "loss": 0.2445, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1283676028251648, + "step": 575, + "valid_targets_mean": 1467.6, + "valid_targets_min": 1009 + }, + { + "epoch": 0.3197353914002205, + "grad_norm": 1.0794408428306252, + "learning_rate": 1.8236220472440946e-05, + "loss": 0.2336, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10993315279483795, + "step": 580, + "valid_targets_mean": 1183.2, + "valid_targets_min": 530 + }, + { + "epoch": 0.3224917309812569, + "grad_norm": 0.9891307523251437, + "learning_rate": 1.8393700787401577e-05, + "loss": 0.2396, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11475048214197159, + "step": 585, + "valid_targets_mean": 1228.0, + "valid_targets_min": 527 + }, + { + "epoch": 0.32524807056229327, + "grad_norm": 0.9690839112503448, + "learning_rate": 1.8551181102362205e-05, + "loss": 0.2384, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12003715336322784, + "step": 590, + "valid_targets_mean": 1422.5, + "valid_targets_min": 1039 + }, + { + "epoch": 0.32800441014332965, + "grad_norm": 1.3809596455033595, + "learning_rate": 1.8708661417322836e-05, + "loss": 0.2411, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10889114439487457, + "step": 595, + "valid_targets_mean": 1187.2, + "valid_targets_min": 601 + }, + { + "epoch": 0.33076074972436603, + "grad_norm": 1.0524564468621989, + "learning_rate": 1.8866141732283464e-05, + "loss": 0.2397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13871678709983826, + "step": 600, + "valid_targets_mean": 1574.2, + "valid_targets_min": 589 + }, + { + "epoch": 0.3335170893054024, + "grad_norm": 0.9932064048778765, + "learning_rate": 1.9023622047244098e-05, + "loss": 0.2387, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13435429334640503, + "step": 605, + "valid_targets_mean": 1395.0, + "valid_targets_min": 659 + }, + { + "epoch": 0.3362734288864388, + "grad_norm": 1.2599368225091614, + "learning_rate": 1.9181102362204726e-05, + "loss": 0.2442, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1276276707649231, + "step": 610, + "valid_targets_mean": 1179.8, + "valid_targets_min": 521 + }, + { + "epoch": 0.3390297684674752, + "grad_norm": 1.112628247450662, + "learning_rate": 1.9338582677165357e-05, + "loss": 0.2417, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06496867537498474, + "step": 615, + "valid_targets_mean": 873.8, + "valid_targets_min": 485 + }, + { + "epoch": 0.34178610804851156, + "grad_norm": 0.9702015273700805, + "learning_rate": 1.9496062992125985e-05, + "loss": 0.2506, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09694847464561462, + "step": 620, + "valid_targets_mean": 1309.2, + "valid_targets_min": 567 + }, + { + "epoch": 0.34454244762954794, + "grad_norm": 0.9223659642751659, + "learning_rate": 1.9653543307086616e-05, + "loss": 0.2511, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14007516205310822, + "step": 625, + "valid_targets_mean": 1689.1, + "valid_targets_min": 558 + }, + { + "epoch": 0.3472987872105843, + "grad_norm": 1.0189945987684093, + "learning_rate": 1.9811023622047244e-05, + "loss": 0.2479, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11825639009475708, + "step": 630, + "valid_targets_mean": 1177.2, + "valid_targets_min": 462 + }, + { + "epoch": 0.3500551267916207, + "grad_norm": 1.018569095169398, + "learning_rate": 1.9968503937007875e-05, + "loss": 0.2396, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.105283722281456, + "step": 635, + "valid_targets_mean": 1081.6, + "valid_targets_min": 526 + }, + { + "epoch": 0.3528114663726571, + "grad_norm": 1.0212192704531378, + "learning_rate": 2.0125984251968503e-05, + "loss": 0.2354, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1239863932132721, + "step": 640, + "valid_targets_mean": 1399.0, + "valid_targets_min": 730 + }, + { + "epoch": 0.35556780595369347, + "grad_norm": 1.1103107724340795, + "learning_rate": 2.0283464566929134e-05, + "loss": 0.235, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12183505296707153, + "step": 645, + "valid_targets_mean": 1232.2, + "valid_targets_min": 591 + }, + { + "epoch": 0.35832414553472985, + "grad_norm": 1.0079759414828307, + "learning_rate": 2.0440944881889762e-05, + "loss": 0.2393, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09770427644252777, + "step": 650, + "valid_targets_mean": 1135.2, + "valid_targets_min": 649 + }, + { + "epoch": 0.36108048511576624, + "grad_norm": 1.1810053632883954, + "learning_rate": 2.0598425196850393e-05, + "loss": 0.229, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08356510102748871, + "step": 655, + "valid_targets_mean": 886.4, + "valid_targets_min": 484 + }, + { + "epoch": 0.3638368246968026, + "grad_norm": 0.9141971228879695, + "learning_rate": 2.0755905511811028e-05, + "loss": 0.2236, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1273081749677658, + "step": 660, + "valid_targets_mean": 1683.6, + "valid_targets_min": 1143 + }, + { + "epoch": 0.36659316427783906, + "grad_norm": 1.0974679704157515, + "learning_rate": 2.0913385826771656e-05, + "loss": 0.2425, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14001387357711792, + "step": 665, + "valid_targets_mean": 1251.6, + "valid_targets_min": 580 + }, + { + "epoch": 0.36934950385887544, + "grad_norm": 0.9590377025410924, + "learning_rate": 2.1070866141732287e-05, + "loss": 0.2332, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10724326223134995, + "step": 670, + "valid_targets_mean": 1225.8, + "valid_targets_min": 698 + }, + { + "epoch": 0.3721058434399118, + "grad_norm": 0.8956761778216485, + "learning_rate": 2.1228346456692915e-05, + "loss": 0.229, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12600812315940857, + "step": 675, + "valid_targets_mean": 1476.8, + "valid_targets_min": 995 + }, + { + "epoch": 0.3748621830209482, + "grad_norm": 0.7177263077715752, + "learning_rate": 2.1385826771653546e-05, + "loss": 0.2226, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09633441269397736, + "step": 680, + "valid_targets_mean": 1516.2, + "valid_targets_min": 724 + }, + { + "epoch": 0.3776185226019846, + "grad_norm": 1.0426251860140405, + "learning_rate": 2.1543307086614177e-05, + "loss": 0.2157, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0979403406381607, + "step": 685, + "valid_targets_mean": 1386.0, + "valid_targets_min": 663 + }, + { + "epoch": 0.38037486218302097, + "grad_norm": 0.9725542335989131, + "learning_rate": 2.1700787401574805e-05, + "loss": 0.2323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1109190583229065, + "step": 690, + "valid_targets_mean": 1444.8, + "valid_targets_min": 1123 + }, + { + "epoch": 0.38313120176405735, + "grad_norm": 0.9316198928506837, + "learning_rate": 2.1858267716535436e-05, + "loss": 0.2339, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11720071732997894, + "step": 695, + "valid_targets_mean": 1276.6, + "valid_targets_min": 613 + }, + { + "epoch": 0.38588754134509373, + "grad_norm": 1.0272538885856828, + "learning_rate": 2.2015748031496064e-05, + "loss": 0.2351, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11373689770698547, + "step": 700, + "valid_targets_mean": 1328.6, + "valid_targets_min": 561 + }, + { + "epoch": 0.3886438809261301, + "grad_norm": 1.0755788665136374, + "learning_rate": 2.2173228346456695e-05, + "loss": 0.2383, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10598930716514587, + "step": 705, + "valid_targets_mean": 1281.2, + "valid_targets_min": 812 + }, + { + "epoch": 0.3914002205071665, + "grad_norm": 1.0420962309628208, + "learning_rate": 2.2330708661417323e-05, + "loss": 0.2276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10730789601802826, + "step": 710, + "valid_targets_mean": 1061.0, + "valid_targets_min": 493 + }, + { + "epoch": 0.3941565600882029, + "grad_norm": 1.0314114407348174, + "learning_rate": 2.2488188976377954e-05, + "loss": 0.2391, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0812050998210907, + "step": 715, + "valid_targets_mean": 999.1, + "valid_targets_min": 502 + }, + { + "epoch": 0.39691289966923926, + "grad_norm": 0.9131925883167099, + "learning_rate": 2.2645669291338582e-05, + "loss": 0.3339, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1989348828792572, + "step": 720, + "valid_targets_mean": 4841.0, + "valid_targets_min": 1966 + }, + { + "epoch": 0.39966923925027564, + "grad_norm": 0.763939641673089, + "learning_rate": 2.2803149606299214e-05, + "loss": 0.4033, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22368791699409485, + "step": 725, + "valid_targets_mean": 4936.8, + "valid_targets_min": 1399 + }, + { + "epoch": 0.402425578831312, + "grad_norm": 0.6319007796899498, + "learning_rate": 2.296062992125984e-05, + "loss": 0.3959, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2510519027709961, + "step": 730, + "valid_targets_mean": 5672.8, + "valid_targets_min": 2678 + }, + { + "epoch": 0.4051819184123484, + "grad_norm": 0.6553504523020409, + "learning_rate": 2.3118110236220473e-05, + "loss": 0.3776, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16071294248104095, + "step": 735, + "valid_targets_mean": 4044.2, + "valid_targets_min": 2322 + }, + { + "epoch": 0.4079382579933848, + "grad_norm": 0.6188291699337207, + "learning_rate": 2.3275590551181107e-05, + "loss": 0.3748, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15347948670387268, + "step": 740, + "valid_targets_mean": 3605.0, + "valid_targets_min": 691 + }, + { + "epoch": 0.41069459757442117, + "grad_norm": 0.6175869837666567, + "learning_rate": 2.3433070866141735e-05, + "loss": 0.3911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14870712161064148, + "step": 745, + "valid_targets_mean": 4230.2, + "valid_targets_min": 921 + }, + { + "epoch": 0.41345093715545755, + "grad_norm": 0.5979216051230694, + "learning_rate": 2.3590551181102366e-05, + "loss": 0.3574, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19179964065551758, + "step": 750, + "valid_targets_mean": 4725.8, + "valid_targets_min": 2127 + }, + { + "epoch": 0.41620727673649394, + "grad_norm": 0.5967006516846378, + "learning_rate": 2.3748031496062994e-05, + "loss": 0.3705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13780179619789124, + "step": 755, + "valid_targets_mean": 2875.8, + "valid_targets_min": 775 + }, + { + "epoch": 0.4189636163175303, + "grad_norm": 0.519010288537669, + "learning_rate": 2.3905511811023625e-05, + "loss": 0.3639, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16765117645263672, + "step": 760, + "valid_targets_mean": 5394.6, + "valid_targets_min": 3997 + }, + { + "epoch": 0.4217199558985667, + "grad_norm": 0.7431657239795751, + "learning_rate": 2.4062992125984253e-05, + "loss": 0.3847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17855000495910645, + "step": 765, + "valid_targets_mean": 3801.0, + "valid_targets_min": 954 + }, + { + "epoch": 0.4244762954796031, + "grad_norm": 0.5477642520128974, + "learning_rate": 2.4220472440944884e-05, + "loss": 0.3646, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1755390465259552, + "step": 770, + "valid_targets_mean": 5662.0, + "valid_targets_min": 717 + }, + { + "epoch": 0.42723263506063947, + "grad_norm": 0.6103477496735357, + "learning_rate": 2.4377952755905516e-05, + "loss": 0.3649, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20465750992298126, + "step": 775, + "valid_targets_mean": 4817.2, + "valid_targets_min": 1109 + }, + { + "epoch": 0.42998897464167585, + "grad_norm": 0.5558753510755993, + "learning_rate": 2.4535433070866143e-05, + "loss": 0.3615, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17041832208633423, + "step": 780, + "valid_targets_mean": 5104.2, + "valid_targets_min": 1374 + }, + { + "epoch": 0.43274531422271223, + "grad_norm": 0.5741922681663364, + "learning_rate": 2.4692913385826775e-05, + "loss": 0.365, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16061919927597046, + "step": 785, + "valid_targets_mean": 3611.5, + "valid_targets_min": 1755 + }, + { + "epoch": 0.4355016538037486, + "grad_norm": 0.7235614270901553, + "learning_rate": 2.4850393700787402e-05, + "loss": 0.3753, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1667340099811554, + "step": 790, + "valid_targets_mean": 2758.9, + "valid_targets_min": 863 + }, + { + "epoch": 0.438257993384785, + "grad_norm": 0.6368910877010386, + "learning_rate": 2.5007874015748034e-05, + "loss": 0.3713, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1953275203704834, + "step": 795, + "valid_targets_mean": 4109.1, + "valid_targets_min": 1229 + }, + { + "epoch": 0.4410143329658214, + "grad_norm": 0.6523565644796165, + "learning_rate": 2.516535433070866e-05, + "loss": 0.3659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19297558069229126, + "step": 800, + "valid_targets_mean": 4061.4, + "valid_targets_min": 2126 + }, + { + "epoch": 0.44377067254685776, + "grad_norm": 0.6666582914006715, + "learning_rate": 2.5322834645669293e-05, + "loss": 0.3784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1980544626712799, + "step": 805, + "valid_targets_mean": 3655.9, + "valid_targets_min": 2053 + }, + { + "epoch": 0.44652701212789414, + "grad_norm": 0.8105642039677585, + "learning_rate": 2.548031496062992e-05, + "loss": 0.3808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20605933666229248, + "step": 810, + "valid_targets_mean": 3319.6, + "valid_targets_min": 630 + }, + { + "epoch": 0.4492833517089305, + "grad_norm": 0.5910538384678853, + "learning_rate": 2.563779527559055e-05, + "loss": 0.3678, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14436432719230652, + "step": 815, + "valid_targets_mean": 3465.5, + "valid_targets_min": 1581 + }, + { + "epoch": 0.4520396912899669, + "grad_norm": 0.6633675876203413, + "learning_rate": 2.5795275590551186e-05, + "loss": 0.3749, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1876346617937088, + "step": 820, + "valid_targets_mean": 3570.1, + "valid_targets_min": 960 + }, + { + "epoch": 0.4547960308710033, + "grad_norm": 0.7339746672497289, + "learning_rate": 2.5952755905511814e-05, + "loss": 0.3521, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21295270323753357, + "step": 825, + "valid_targets_mean": 4040.0, + "valid_targets_min": 2297 + }, + { + "epoch": 0.45755237045203967, + "grad_norm": 0.645406766930925, + "learning_rate": 2.6110236220472445e-05, + "loss": 0.3494, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19218125939369202, + "step": 830, + "valid_targets_mean": 4187.5, + "valid_targets_min": 1749 + }, + { + "epoch": 0.46030871003307605, + "grad_norm": 0.7644646305275609, + "learning_rate": 2.6267716535433073e-05, + "loss": 0.3512, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23554390668869019, + "step": 835, + "valid_targets_mean": 4211.1, + "valid_targets_min": 2076 + }, + { + "epoch": 0.46306504961411243, + "grad_norm": 0.6661506734652453, + "learning_rate": 2.6425196850393704e-05, + "loss": 0.3652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17666974663734436, + "step": 840, + "valid_targets_mean": 3433.4, + "valid_targets_min": 1743 + }, + { + "epoch": 0.4658213891951488, + "grad_norm": 0.6936097398400373, + "learning_rate": 2.6582677165354332e-05, + "loss": 0.3658, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16362440586090088, + "step": 845, + "valid_targets_mean": 3017.8, + "valid_targets_min": 1043 + }, + { + "epoch": 0.4685777287761852, + "grad_norm": 0.6410477849033325, + "learning_rate": 2.6740157480314963e-05, + "loss": 0.3456, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16146893799304962, + "step": 850, + "valid_targets_mean": 4154.0, + "valid_targets_min": 956 + }, + { + "epoch": 0.47133406835722164, + "grad_norm": 0.5462424333871169, + "learning_rate": 2.689763779527559e-05, + "loss": 0.348, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14395523071289062, + "step": 855, + "valid_targets_mean": 4138.5, + "valid_targets_min": 1389 + }, + { + "epoch": 0.474090407938258, + "grad_norm": 0.6543806324321705, + "learning_rate": 2.7055118110236222e-05, + "loss": 0.359, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15437737107276917, + "step": 860, + "valid_targets_mean": 3197.0, + "valid_targets_min": 1997 + }, + { + "epoch": 0.4768467475192944, + "grad_norm": 0.6275742283275833, + "learning_rate": 2.7212598425196854e-05, + "loss": 0.3569, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16257920861244202, + "step": 865, + "valid_targets_mean": 3766.9, + "valid_targets_min": 2132 + }, + { + "epoch": 0.4796030871003308, + "grad_norm": 0.6501666094252877, + "learning_rate": 2.737007874015748e-05, + "loss": 0.3445, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1705726683139801, + "step": 870, + "valid_targets_mean": 3891.9, + "valid_targets_min": 768 + }, + { + "epoch": 0.48235942668136716, + "grad_norm": 0.6902479128736246, + "learning_rate": 2.7527559055118113e-05, + "loss": 0.3472, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12727327644824982, + "step": 875, + "valid_targets_mean": 2598.5, + "valid_targets_min": 940 + }, + { + "epoch": 0.48511576626240355, + "grad_norm": 0.7725484755750387, + "learning_rate": 2.768503937007874e-05, + "loss": 0.359, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14408493041992188, + "step": 880, + "valid_targets_mean": 2420.8, + "valid_targets_min": 1118 + }, + { + "epoch": 0.48787210584343993, + "grad_norm": 0.6391554760311697, + "learning_rate": 2.7842519685039372e-05, + "loss": 0.337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15387314558029175, + "step": 885, + "valid_targets_mean": 3099.2, + "valid_targets_min": 1541 + }, + { + "epoch": 0.4906284454244763, + "grad_norm": 0.804647382033559, + "learning_rate": 2.8e-05, + "loss": 0.3648, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20953971147537231, + "step": 890, + "valid_targets_mean": 2600.6, + "valid_targets_min": 617 + }, + { + "epoch": 0.4933847850055127, + "grad_norm": 0.8695786713335164, + "learning_rate": 2.815748031496063e-05, + "loss": 0.3568, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19551296532154083, + "step": 895, + "valid_targets_mean": 2638.8, + "valid_targets_min": 888 + }, + { + "epoch": 0.4961411245865491, + "grad_norm": 0.6586371298688602, + "learning_rate": 2.8314960629921265e-05, + "loss": 0.3506, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18856769800186157, + "step": 900, + "valid_targets_mean": 3712.1, + "valid_targets_min": 2268 + }, + { + "epoch": 0.49889746416758546, + "grad_norm": 0.6745323293954485, + "learning_rate": 2.8472440944881893e-05, + "loss": 0.3427, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16712594032287598, + "step": 905, + "valid_targets_mean": 2838.0, + "valid_targets_min": 1840 + }, + { + "epoch": 0.5016538037486218, + "grad_norm": 0.7056683080325208, + "learning_rate": 2.8629921259842524e-05, + "loss": 0.3448, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14798742532730103, + "step": 910, + "valid_targets_mean": 2626.9, + "valid_targets_min": 761 + }, + { + "epoch": 0.5044101433296582, + "grad_norm": 0.6613055586176625, + "learning_rate": 2.8787401574803152e-05, + "loss": 0.3404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1502208411693573, + "step": 915, + "valid_targets_mean": 2880.5, + "valid_targets_min": 826 + }, + { + "epoch": 0.5071664829106945, + "grad_norm": 0.6679534271210537, + "learning_rate": 2.8944881889763783e-05, + "loss": 0.351, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16233298182487488, + "step": 920, + "valid_targets_mean": 3267.4, + "valid_targets_min": 1727 + }, + { + "epoch": 0.5099228224917309, + "grad_norm": 0.610163233853472, + "learning_rate": 2.910236220472441e-05, + "loss": 0.3504, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14249645173549652, + "step": 925, + "valid_targets_mean": 3612.2, + "valid_targets_min": 989 + }, + { + "epoch": 0.5126791620727673, + "grad_norm": 0.6498589346686571, + "learning_rate": 2.9259842519685043e-05, + "loss": 0.3477, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16696658730506897, + "step": 930, + "valid_targets_mean": 3919.2, + "valid_targets_min": 1076 + }, + { + "epoch": 0.5154355016538037, + "grad_norm": 0.6241007377245298, + "learning_rate": 2.941732283464567e-05, + "loss": 0.344, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19479955732822418, + "step": 935, + "valid_targets_mean": 4062.1, + "valid_targets_min": 1537 + }, + { + "epoch": 0.5181918412348401, + "grad_norm": 0.6110745770696611, + "learning_rate": 2.95748031496063e-05, + "loss": 0.3371, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19176682829856873, + "step": 940, + "valid_targets_mean": 4059.4, + "valid_targets_min": 2065 + }, + { + "epoch": 0.5209481808158766, + "grad_norm": 0.815342354863853, + "learning_rate": 2.9732283464566933e-05, + "loss": 0.3466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17000389099121094, + "step": 945, + "valid_targets_mean": 3896.6, + "valid_targets_min": 2047 + }, + { + "epoch": 0.523704520396913, + "grad_norm": 0.593075525803173, + "learning_rate": 2.988976377952756e-05, + "loss": 0.3388, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18321101367473602, + "step": 950, + "valid_targets_mean": 3880.2, + "valid_targets_min": 1307 + }, + { + "epoch": 0.5264608599779493, + "grad_norm": 0.7311196196960724, + "learning_rate": 3.0047244094488192e-05, + "loss": 0.3554, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1731484830379486, + "step": 955, + "valid_targets_mean": 2856.6, + "valid_targets_min": 807 + }, + { + "epoch": 0.5292171995589857, + "grad_norm": 0.6207253050685397, + "learning_rate": 3.020472440944882e-05, + "loss": 0.3562, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14484131336212158, + "step": 960, + "valid_targets_mean": 2880.9, + "valid_targets_min": 1201 + }, + { + "epoch": 0.5319735391400221, + "grad_norm": 0.604095081633314, + "learning_rate": 3.036220472440945e-05, + "loss": 0.338, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1396373212337494, + "step": 965, + "valid_targets_mean": 3004.5, + "valid_targets_min": 300 + }, + { + "epoch": 0.5347298787210585, + "grad_norm": 0.618969855136168, + "learning_rate": 3.051968503937008e-05, + "loss": 0.3415, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1660369634628296, + "step": 970, + "valid_targets_mean": 3600.9, + "valid_targets_min": 2462 + }, + { + "epoch": 0.5374862183020949, + "grad_norm": 0.6471484368375467, + "learning_rate": 3.067716535433071e-05, + "loss": 0.3464, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18496090173721313, + "step": 975, + "valid_targets_mean": 4155.0, + "valid_targets_min": 1227 + }, + { + "epoch": 0.5402425578831312, + "grad_norm": 0.642161873290024, + "learning_rate": 3.0834645669291345e-05, + "loss": 0.343, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14892789721488953, + "step": 980, + "valid_targets_mean": 3017.0, + "valid_targets_min": 964 + }, + { + "epoch": 0.5429988974641676, + "grad_norm": 0.6526116765116071, + "learning_rate": 3.099212598425197e-05, + "loss": 0.3218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17063620686531067, + "step": 985, + "valid_targets_mean": 3252.9, + "valid_targets_min": 2045 + }, + { + "epoch": 0.545755237045204, + "grad_norm": 0.6099416180153985, + "learning_rate": 3.11496062992126e-05, + "loss": 0.3356, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18031102418899536, + "step": 990, + "valid_targets_mean": 4026.8, + "valid_targets_min": 840 + }, + { + "epoch": 0.5485115766262404, + "grad_norm": 0.6887845602807509, + "learning_rate": 3.1307086614173235e-05, + "loss": 0.3156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17059743404388428, + "step": 995, + "valid_targets_mean": 4145.4, + "valid_targets_min": 512 + }, + { + "epoch": 0.5512679162072768, + "grad_norm": 0.629766598340137, + "learning_rate": 3.146456692913386e-05, + "loss": 0.3608, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1923399418592453, + "step": 1000, + "valid_targets_mean": 3610.0, + "valid_targets_min": 1651 + }, + { + "epoch": 0.5540242557883132, + "grad_norm": 0.600868128092, + "learning_rate": 3.162204724409449e-05, + "loss": 0.332, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1322053074836731, + "step": 1005, + "valid_targets_mean": 2796.1, + "valid_targets_min": 1739 + }, + { + "epoch": 0.5567805953693495, + "grad_norm": 0.6666928885711653, + "learning_rate": 3.177952755905512e-05, + "loss": 0.3434, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17031845450401306, + "step": 1010, + "valid_targets_mean": 3076.9, + "valid_targets_min": 594 + }, + { + "epoch": 0.5595369349503859, + "grad_norm": 0.6491351383518987, + "learning_rate": 3.193700787401575e-05, + "loss": 0.3432, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17624753713607788, + "step": 1015, + "valid_targets_mean": 3337.9, + "valid_targets_min": 1566 + }, + { + "epoch": 0.5622932745314223, + "grad_norm": 0.6044789041237325, + "learning_rate": 3.209448818897638e-05, + "loss": 0.3497, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18272504210472107, + "step": 1020, + "valid_targets_mean": 3851.8, + "valid_targets_min": 1213 + }, + { + "epoch": 0.5650496141124587, + "grad_norm": 0.6792766889837897, + "learning_rate": 3.225196850393701e-05, + "loss": 0.3421, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1880827248096466, + "step": 1025, + "valid_targets_mean": 2973.0, + "valid_targets_min": 1215 + }, + { + "epoch": 0.5678059536934951, + "grad_norm": 0.5810954986197091, + "learning_rate": 3.240944881889764e-05, + "loss": 0.3221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1828366369009018, + "step": 1030, + "valid_targets_mean": 4297.1, + "valid_targets_min": 2716 + }, + { + "epoch": 0.5705622932745315, + "grad_norm": 0.620551072714132, + "learning_rate": 3.256692913385827e-05, + "loss": 0.3317, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1770530492067337, + "step": 1035, + "valid_targets_mean": 3919.9, + "valid_targets_min": 1955 + }, + { + "epoch": 0.5733186328555678, + "grad_norm": 0.7099487439178426, + "learning_rate": 3.27244094488189e-05, + "loss": 0.3334, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17522679269313812, + "step": 1040, + "valid_targets_mean": 3298.9, + "valid_targets_min": 643 + }, + { + "epoch": 0.5760749724366042, + "grad_norm": 0.6450009748970515, + "learning_rate": 3.2881889763779527e-05, + "loss": 0.3258, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1641930341720581, + "step": 1045, + "valid_targets_mean": 3446.2, + "valid_targets_min": 966 + }, + { + "epoch": 0.5788313120176406, + "grad_norm": 0.6563631367095679, + "learning_rate": 3.303937007874016e-05, + "loss": 0.3111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1406809538602829, + "step": 1050, + "valid_targets_mean": 2767.8, + "valid_targets_min": 1101 + }, + { + "epoch": 0.581587651598677, + "grad_norm": 0.6629156824123476, + "learning_rate": 3.319685039370079e-05, + "loss": 0.3233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15934285521507263, + "step": 1055, + "valid_targets_mean": 2842.2, + "valid_targets_min": 610 + }, + { + "epoch": 0.5843439911797134, + "grad_norm": 0.678485158984605, + "learning_rate": 3.3354330708661424e-05, + "loss": 0.3352, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19007810950279236, + "step": 1060, + "valid_targets_mean": 3791.8, + "valid_targets_min": 1285 + }, + { + "epoch": 0.5871003307607497, + "grad_norm": 0.6459233036864768, + "learning_rate": 3.351181102362205e-05, + "loss": 0.3592, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15802001953125, + "step": 1065, + "valid_targets_mean": 3571.8, + "valid_targets_min": 1236 + }, + { + "epoch": 0.5898566703417861, + "grad_norm": 0.611548840094281, + "learning_rate": 3.366929133858268e-05, + "loss": 0.3285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18245014548301697, + "step": 1070, + "valid_targets_mean": 3953.5, + "valid_targets_min": 2520 + }, + { + "epoch": 0.5926130099228225, + "grad_norm": 0.6250842637617516, + "learning_rate": 3.3826771653543314e-05, + "loss": 0.3426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17527180910110474, + "step": 1075, + "valid_targets_mean": 3936.1, + "valid_targets_min": 1136 + }, + { + "epoch": 0.5953693495038589, + "grad_norm": 0.6491629982279236, + "learning_rate": 3.398425196850394e-05, + "loss": 0.3211, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17461879551410675, + "step": 1080, + "valid_targets_mean": 3288.1, + "valid_targets_min": 1728 + }, + { + "epoch": 0.5981256890848953, + "grad_norm": 0.6711956331122507, + "learning_rate": 3.414173228346457e-05, + "loss": 0.3337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14062397181987762, + "step": 1085, + "valid_targets_mean": 3756.9, + "valid_targets_min": 1631 + }, + { + "epoch": 0.6008820286659317, + "grad_norm": 0.6105705295036294, + "learning_rate": 3.42992125984252e-05, + "loss": 0.3269, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14773385226726532, + "step": 1090, + "valid_targets_mean": 3669.8, + "valid_targets_min": 2408 + }, + { + "epoch": 0.603638368246968, + "grad_norm": 0.5987555631300199, + "learning_rate": 3.445669291338583e-05, + "loss": 0.3103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1756279021501541, + "step": 1095, + "valid_targets_mean": 3905.9, + "valid_targets_min": 1763 + }, + { + "epoch": 0.6063947078280044, + "grad_norm": 0.6741832609743295, + "learning_rate": 3.461417322834646e-05, + "loss": 0.3439, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19677454233169556, + "step": 1100, + "valid_targets_mean": 2997.2, + "valid_targets_min": 1643 + }, + { + "epoch": 0.6091510474090408, + "grad_norm": 0.6965276948442168, + "learning_rate": 3.477165354330709e-05, + "loss": 0.3255, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14498120546340942, + "step": 1105, + "valid_targets_mean": 2774.0, + "valid_targets_min": 1273 + }, + { + "epoch": 0.6119073869900772, + "grad_norm": 0.8599412878393429, + "learning_rate": 3.4929133858267715e-05, + "loss": 0.3413, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15961644053459167, + "step": 1110, + "valid_targets_mean": 2949.0, + "valid_targets_min": 770 + }, + { + "epoch": 0.6146637265711136, + "grad_norm": 0.7412295650239527, + "learning_rate": 3.508661417322835e-05, + "loss": 0.3403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14571735262870789, + "step": 1115, + "valid_targets_mean": 2409.8, + "valid_targets_min": 364 + }, + { + "epoch": 0.61742006615215, + "grad_norm": 0.562340579409034, + "learning_rate": 3.524409448818898e-05, + "loss": 0.3335, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1772332489490509, + "step": 1120, + "valid_targets_mean": 4212.2, + "valid_targets_min": 1168 + }, + { + "epoch": 0.6201764057331863, + "grad_norm": 0.6899337600171722, + "learning_rate": 3.5401574803149606e-05, + "loss": 0.3699, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19966909289360046, + "step": 1125, + "valid_targets_mean": 3440.5, + "valid_targets_min": 1560 + }, + { + "epoch": 0.6229327453142227, + "grad_norm": 0.6581835099516995, + "learning_rate": 3.555905511811024e-05, + "loss": 0.3329, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15129685401916504, + "step": 1130, + "valid_targets_mean": 2950.2, + "valid_targets_min": 1373 + }, + { + "epoch": 0.6256890848952591, + "grad_norm": 0.6043419538977272, + "learning_rate": 3.571653543307087e-05, + "loss": 0.3416, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13217024505138397, + "step": 1135, + "valid_targets_mean": 3080.8, + "valid_targets_min": 1435 + }, + { + "epoch": 0.6284454244762955, + "grad_norm": 0.6506299537631127, + "learning_rate": 3.58740157480315e-05, + "loss": 0.3435, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2069811224937439, + "step": 1140, + "valid_targets_mean": 4122.6, + "valid_targets_min": 2204 + }, + { + "epoch": 0.6312017640573319, + "grad_norm": 0.5704958592291968, + "learning_rate": 3.603149606299213e-05, + "loss": 0.3394, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19020536541938782, + "step": 1145, + "valid_targets_mean": 5108.1, + "valid_targets_min": 2223 + }, + { + "epoch": 0.6339581036383682, + "grad_norm": 0.602894961360399, + "learning_rate": 3.618897637795276e-05, + "loss": 0.3396, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19966936111450195, + "step": 1150, + "valid_targets_mean": 4091.5, + "valid_targets_min": 1137 + }, + { + "epoch": 0.6367144432194046, + "grad_norm": 0.6296413268659747, + "learning_rate": 3.634645669291339e-05, + "loss": 0.3377, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13153016567230225, + "step": 1155, + "valid_targets_mean": 3131.0, + "valid_targets_min": 373 + }, + { + "epoch": 0.639470782800441, + "grad_norm": 0.6468244359742638, + "learning_rate": 3.650393700787402e-05, + "loss": 0.3379, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1477544754743576, + "step": 1160, + "valid_targets_mean": 2893.9, + "valid_targets_min": 352 + }, + { + "epoch": 0.6422271223814774, + "grad_norm": 0.5556879705210325, + "learning_rate": 3.666141732283465e-05, + "loss": 0.3375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14806342124938965, + "step": 1165, + "valid_targets_mean": 4443.4, + "valid_targets_min": 2838 + }, + { + "epoch": 0.6449834619625138, + "grad_norm": 0.9443667257374213, + "learning_rate": 3.6818897637795276e-05, + "loss": 0.3282, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1718931794166565, + "step": 1170, + "valid_targets_mean": 3518.8, + "valid_targets_min": 1369 + }, + { + "epoch": 0.6477398015435502, + "grad_norm": 0.6199386836136143, + "learning_rate": 3.697637795275591e-05, + "loss": 0.3141, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19095441699028015, + "step": 1175, + "valid_targets_mean": 4071.6, + "valid_targets_min": 2579 + }, + { + "epoch": 0.6504961411245865, + "grad_norm": 0.6036105120126052, + "learning_rate": 3.713385826771654e-05, + "loss": 0.3296, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12086024135351181, + "step": 1180, + "valid_targets_mean": 2293.9, + "valid_targets_min": 306 + }, + { + "epoch": 0.6532524807056229, + "grad_norm": 0.6117833610460935, + "learning_rate": 3.729133858267717e-05, + "loss": 0.321, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15778464078903198, + "step": 1185, + "valid_targets_mean": 3565.0, + "valid_targets_min": 1399 + }, + { + "epoch": 0.6560088202866593, + "grad_norm": 0.6130497172678445, + "learning_rate": 3.7448818897637795e-05, + "loss": 0.3312, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1479160487651825, + "step": 1190, + "valid_targets_mean": 3098.2, + "valid_targets_min": 1062 + }, + { + "epoch": 0.6587651598676957, + "grad_norm": 0.7202734772982229, + "learning_rate": 3.760629921259843e-05, + "loss": 0.3309, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18763455748558044, + "step": 1195, + "valid_targets_mean": 3322.4, + "valid_targets_min": 1093 + }, + { + "epoch": 0.6615214994487321, + "grad_norm": 0.5803590942587097, + "learning_rate": 3.776377952755906e-05, + "loss": 0.3286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16870462894439697, + "step": 1200, + "valid_targets_mean": 3481.9, + "valid_targets_min": 313 + }, + { + "epoch": 0.6642778390297684, + "grad_norm": 0.9297196675788153, + "learning_rate": 3.7921259842519685e-05, + "loss": 0.3349, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16214972734451294, + "step": 1205, + "valid_targets_mean": 2449.4, + "valid_targets_min": 783 + }, + { + "epoch": 0.6670341786108048, + "grad_norm": 0.6130140277189999, + "learning_rate": 3.807874015748032e-05, + "loss": 0.3297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19512765109539032, + "step": 1210, + "valid_targets_mean": 4458.8, + "valid_targets_min": 1177 + }, + { + "epoch": 0.6697905181918412, + "grad_norm": 0.5609881653518721, + "learning_rate": 3.823622047244095e-05, + "loss": 0.3181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12719246745109558, + "step": 1215, + "valid_targets_mean": 3365.0, + "valid_targets_min": 1972 + }, + { + "epoch": 0.6725468577728776, + "grad_norm": 0.6146555098166758, + "learning_rate": 3.839370078740158e-05, + "loss": 0.3112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1824508160352707, + "step": 1220, + "valid_targets_mean": 4161.5, + "valid_targets_min": 2550 + }, + { + "epoch": 0.675303197353914, + "grad_norm": 0.48618418512877476, + "learning_rate": 3.855118110236221e-05, + "loss": 0.3147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15225398540496826, + "step": 1225, + "valid_targets_mean": 6125.4, + "valid_targets_min": 2132 + }, + { + "epoch": 0.6780595369349504, + "grad_norm": 0.5614097814518544, + "learning_rate": 3.870866141732284e-05, + "loss": 0.3242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16556255519390106, + "step": 1230, + "valid_targets_mean": 3607.5, + "valid_targets_min": 2280 + }, + { + "epoch": 0.6808158765159867, + "grad_norm": 0.6414300716172275, + "learning_rate": 3.886614173228347e-05, + "loss": 0.3134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1355023831129074, + "step": 1235, + "valid_targets_mean": 2645.5, + "valid_targets_min": 794 + }, + { + "epoch": 0.6835722160970231, + "grad_norm": 0.6157167962571859, + "learning_rate": 3.90236220472441e-05, + "loss": 0.3152, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15326917171478271, + "step": 1240, + "valid_targets_mean": 3561.2, + "valid_targets_min": 1293 + }, + { + "epoch": 0.6863285556780595, + "grad_norm": 0.6181905285662852, + "learning_rate": 3.918110236220473e-05, + "loss": 0.3196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1768895834684372, + "step": 1245, + "valid_targets_mean": 3339.4, + "valid_targets_min": 1357 + }, + { + "epoch": 0.6890848952590959, + "grad_norm": 0.591432816157219, + "learning_rate": 3.9338582677165356e-05, + "loss": 0.3152, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14954379200935364, + "step": 1250, + "valid_targets_mean": 3513.9, + "valid_targets_min": 2360 + }, + { + "epoch": 0.6918412348401323, + "grad_norm": 0.6179933346934904, + "learning_rate": 3.949606299212599e-05, + "loss": 0.3443, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18894536793231964, + "step": 1255, + "valid_targets_mean": 3954.4, + "valid_targets_min": 1213 + }, + { + "epoch": 0.6945975744211687, + "grad_norm": 0.6461688481796284, + "learning_rate": 3.965354330708662e-05, + "loss": 0.3306, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17518353462219238, + "step": 1260, + "valid_targets_mean": 3356.9, + "valid_targets_min": 1708 + }, + { + "epoch": 0.697353914002205, + "grad_norm": 0.6875902170692092, + "learning_rate": 3.9811023622047246e-05, + "loss": 0.3329, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1744455248117447, + "step": 1265, + "valid_targets_mean": 2848.6, + "valid_targets_min": 1242 + }, + { + "epoch": 0.7001102535832414, + "grad_norm": 0.659456756315284, + "learning_rate": 3.9968503937007874e-05, + "loss": 0.3101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17162586748600006, + "step": 1270, + "valid_targets_mean": 2673.5, + "valid_targets_min": 750 + }, + { + "epoch": 0.7028665931642778, + "grad_norm": 0.6579924894365851, + "learning_rate": 3.9999987908526714e-05, + "loss": 0.3257, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12482356280088425, + "step": 1275, + "valid_targets_mean": 2704.1, + "valid_targets_min": 280 + }, + { + "epoch": 0.7056229327453142, + "grad_norm": 0.6575607373601394, + "learning_rate": 3.999993878694153e-05, + "loss": 0.3202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18606334924697876, + "step": 1280, + "valid_targets_mean": 2230.9, + "valid_targets_min": 313 + }, + { + "epoch": 0.7083792723263506, + "grad_norm": 0.5810230705734479, + "learning_rate": 3.99998518796201e-05, + "loss": 0.3306, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1467871069908142, + "step": 1285, + "valid_targets_mean": 3992.8, + "valid_targets_min": 2193 + }, + { + "epoch": 0.7111356119073869, + "grad_norm": 0.6255006506162679, + "learning_rate": 3.9999727186726634e-05, + "loss": 0.3234, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17592358589172363, + "step": 1290, + "valid_targets_mean": 3137.5, + "valid_targets_min": 325 + }, + { + "epoch": 0.7138919514884233, + "grad_norm": 0.6038134255275343, + "learning_rate": 3.9999564708496694e-05, + "loss": 0.3201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14182081818580627, + "step": 1295, + "valid_targets_mean": 3164.1, + "valid_targets_min": 1286 + }, + { + "epoch": 0.7166482910694597, + "grad_norm": 0.6260652760391379, + "learning_rate": 3.9999364445237255e-05, + "loss": 0.3324, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18128405511379242, + "step": 1300, + "valid_targets_mean": 3985.8, + "valid_targets_min": 1765 + }, + { + "epoch": 0.7194046306504961, + "grad_norm": 0.8053831134942181, + "learning_rate": 3.999912639732668e-05, + "loss": 0.2894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0894903615117073, + "step": 1305, + "valid_targets_mean": 2390.1, + "valid_targets_min": 872 + }, + { + "epoch": 0.7221609702315325, + "grad_norm": 0.6512851870755102, + "learning_rate": 3.99988505652147e-05, + "loss": 0.2016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07160298526287079, + "step": 1310, + "valid_targets_mean": 1543.9, + "valid_targets_min": 665 + }, + { + "epoch": 0.7249173098125689, + "grad_norm": 0.5749917364554537, + "learning_rate": 3.999853694942246e-05, + "loss": 0.1539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07964788377285004, + "step": 1315, + "valid_targets_mean": 3129.8, + "valid_targets_min": 2254 + }, + { + "epoch": 0.7276736493936052, + "grad_norm": 0.5137959951283632, + "learning_rate": 3.999818555054245e-05, + "loss": 0.148, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07077719271183014, + "step": 1320, + "valid_targets_mean": 3026.8, + "valid_targets_min": 706 + }, + { + "epoch": 0.7304299889746417, + "grad_norm": 0.5994561079039828, + "learning_rate": 3.999779636923858e-05, + "loss": 0.1594, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0476398840546608, + "step": 1325, + "valid_targets_mean": 3207.8, + "valid_targets_min": 1216 + }, + { + "epoch": 0.7331863285556781, + "grad_norm": 0.5128209868640025, + "learning_rate": 3.9997369406246115e-05, + "loss": 0.148, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08894810080528259, + "step": 1330, + "valid_targets_mean": 3540.0, + "valid_targets_min": 2551 + }, + { + "epoch": 0.7359426681367145, + "grad_norm": 0.8736863713975318, + "learning_rate": 3.9996904662371736e-05, + "loss": 0.2039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09230896830558777, + "step": 1335, + "valid_targets_mean": 960.9, + "valid_targets_min": 555 + }, + { + "epoch": 0.7386990077177509, + "grad_norm": 0.426155709179169, + "learning_rate": 3.999640213849345e-05, + "loss": 0.1894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06374624371528625, + "step": 1340, + "valid_targets_mean": 3465.4, + "valid_targets_min": 1008 + }, + { + "epoch": 0.7414553472987873, + "grad_norm": 0.5089770186274714, + "learning_rate": 3.99958618355607e-05, + "loss": 0.1613, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08072920143604279, + "step": 1345, + "valid_targets_mean": 3522.1, + "valid_targets_min": 3039 + }, + { + "epoch": 0.7442116868798236, + "grad_norm": 0.5570020109192678, + "learning_rate": 3.9995283754594256e-05, + "loss": 0.2013, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11707267165184021, + "step": 1350, + "valid_targets_mean": 3348.0, + "valid_targets_min": 1938 + }, + { + "epoch": 0.74696802646086, + "grad_norm": 0.3578253695965249, + "learning_rate": 3.999466789668629e-05, + "loss": 0.122, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06903842836618423, + "step": 1355, + "valid_targets_mean": 3973.6, + "valid_targets_min": 924 + }, + { + "epoch": 0.7497243660418964, + "grad_norm": 0.5671628658095135, + "learning_rate": 3.999401426300034e-05, + "loss": 0.1421, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07003912329673767, + "step": 1360, + "valid_targets_mean": 2262.8, + "valid_targets_min": 444 + }, + { + "epoch": 0.7524807056229328, + "grad_norm": 0.4262304976850719, + "learning_rate": 3.99933228547713e-05, + "loss": 0.1388, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0667363777756691, + "step": 1365, + "valid_targets_mean": 3283.2, + "valid_targets_min": 749 + }, + { + "epoch": 0.7552370452039692, + "grad_norm": 0.6377274584274807, + "learning_rate": 3.999259367330546e-05, + "loss": 0.1703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0806315690279007, + "step": 1370, + "valid_targets_mean": 1381.4, + "valid_targets_min": 601 + }, + { + "epoch": 0.7579933847850056, + "grad_norm": 0.4918702441179698, + "learning_rate": 3.999182671998044e-05, + "loss": 0.1591, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07424218952655792, + "step": 1375, + "valid_targets_mean": 3375.0, + "valid_targets_min": 1756 + }, + { + "epoch": 0.7607497243660419, + "grad_norm": 0.5579148305283752, + "learning_rate": 3.9991021996245256e-05, + "loss": 0.1419, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06456337869167328, + "step": 1380, + "valid_targets_mean": 2760.6, + "valid_targets_min": 571 + }, + { + "epoch": 0.7635060639470783, + "grad_norm": 0.54289923920405, + "learning_rate": 3.9990179503620255e-05, + "loss": 0.1329, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0666951984167099, + "step": 1385, + "valid_targets_mean": 2584.8, + "valid_targets_min": 520 + }, + { + "epoch": 0.7662624035281147, + "grad_norm": 0.5432479876586153, + "learning_rate": 3.9989299243697156e-05, + "loss": 0.134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09289132058620453, + "step": 1390, + "valid_targets_mean": 3460.2, + "valid_targets_min": 2179 + }, + { + "epoch": 0.7690187431091511, + "grad_norm": 0.5926743016063887, + "learning_rate": 3.9988381218139024e-05, + "loss": 0.1457, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059315793216228485, + "step": 1395, + "valid_targets_mean": 3006.9, + "valid_targets_min": 807 + }, + { + "epoch": 0.7717750826901875, + "grad_norm": 0.4593582681600235, + "learning_rate": 3.998742542868028e-05, + "loss": 0.154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06242910400032997, + "step": 1400, + "valid_targets_mean": 1811.0, + "valid_targets_min": 594 + }, + { + "epoch": 0.7745314222712238, + "grad_norm": 0.5431419350802359, + "learning_rate": 3.998643187712669e-05, + "loss": 0.1578, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07660562545061111, + "step": 1405, + "valid_targets_mean": 2210.0, + "valid_targets_min": 714 + }, + { + "epoch": 0.7772877618522602, + "grad_norm": 0.5675126983836004, + "learning_rate": 3.9985400565355365e-05, + "loss": 0.175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14711914956569672, + "step": 1410, + "valid_targets_mean": 3098.4, + "valid_targets_min": 1227 + }, + { + "epoch": 0.7800441014332966, + "grad_norm": 0.4902598847908753, + "learning_rate": 3.998433149531475e-05, + "loss": 0.1555, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06333103775978088, + "step": 1415, + "valid_targets_mean": 2710.6, + "valid_targets_min": 782 + }, + { + "epoch": 0.782800441014333, + "grad_norm": 0.73967666065278, + "learning_rate": 3.998322466902464e-05, + "loss": 0.1385, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10648548603057861, + "step": 1420, + "valid_targets_mean": 1954.5, + "valid_targets_min": 950 + }, + { + "epoch": 0.7855567805953694, + "grad_norm": 0.46961432753237214, + "learning_rate": 3.998208008857614e-05, + "loss": 0.1855, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07578346133232117, + "step": 1425, + "valid_targets_mean": 3065.0, + "valid_targets_min": 1296 + }, + { + "epoch": 0.7883131201764058, + "grad_norm": 0.45130081883954204, + "learning_rate": 3.9980897756131716e-05, + "loss": 0.1634, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053175389766693115, + "step": 1430, + "valid_targets_mean": 3146.6, + "valid_targets_min": 789 + }, + { + "epoch": 0.7910694597574421, + "grad_norm": 0.6040878995125983, + "learning_rate": 3.9979677673925125e-05, + "loss": 0.1453, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08629291504621506, + "step": 1435, + "valid_targets_mean": 2301.5, + "valid_targets_min": 713 + }, + { + "epoch": 0.7938257993384785, + "grad_norm": 0.5630625539469728, + "learning_rate": 3.997841984426147e-05, + "loss": 0.3529, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15099017322063446, + "step": 1440, + "valid_targets_mean": 2630.1, + "valid_targets_min": 980 + }, + { + "epoch": 0.7965821389195149, + "grad_norm": 0.7231222219798372, + "learning_rate": 3.997712426951715e-05, + "loss": 0.1671, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1257159411907196, + "step": 1445, + "valid_targets_mean": 1792.1, + "valid_targets_min": 753 + }, + { + "epoch": 0.7993384785005513, + "grad_norm": 0.6047357929048681, + "learning_rate": 3.997579095213988e-05, + "loss": 0.1454, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05913805589079857, + "step": 1450, + "valid_targets_mean": 1459.5, + "valid_targets_min": 670 + }, + { + "epoch": 0.8020948180815877, + "grad_norm": 0.8619410797918173, + "learning_rate": 3.997441989464871e-05, + "loss": 0.2544, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08314186334609985, + "step": 1455, + "valid_targets_mean": 2546.0, + "valid_targets_min": 663 + }, + { + "epoch": 0.804851157662624, + "grad_norm": 0.31416614610257443, + "learning_rate": 3.997301109963393e-05, + "loss": 0.1253, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039348430931568146, + "step": 1460, + "valid_targets_mean": 5186.2, + "valid_targets_min": 3522 + }, + { + "epoch": 0.8076074972436604, + "grad_norm": 0.45878735313807617, + "learning_rate": 3.997156456975721e-05, + "loss": 0.1419, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08301390707492828, + "step": 1465, + "valid_targets_mean": 4087.4, + "valid_targets_min": 2327 + }, + { + "epoch": 0.8103638368246968, + "grad_norm": 0.4260993799083031, + "learning_rate": 3.997008030775144e-05, + "loss": 0.1399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07315197587013245, + "step": 1470, + "valid_targets_mean": 3396.5, + "valid_targets_min": 1218 + }, + { + "epoch": 0.8131201764057332, + "grad_norm": 0.4156062461432017, + "learning_rate": 3.9968558316420834e-05, + "loss": 0.1324, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07940724492073059, + "step": 1475, + "valid_targets_mean": 4247.9, + "valid_targets_min": 3261 + }, + { + "epoch": 0.8158765159867696, + "grad_norm": 0.45155163785817787, + "learning_rate": 3.996699859864088e-05, + "loss": 0.1302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06301572918891907, + "step": 1480, + "valid_targets_mean": 2142.2, + "valid_targets_min": 719 + }, + { + "epoch": 0.818632855567806, + "grad_norm": 0.45706836958294494, + "learning_rate": 3.996540115735833e-05, + "loss": 0.1291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0699104517698288, + "step": 1485, + "valid_targets_mean": 3404.1, + "valid_targets_min": 2224 + }, + { + "epoch": 0.8213891951488423, + "grad_norm": 0.3881661216221874, + "learning_rate": 3.996376599559124e-05, + "loss": 0.1231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04443890601396561, + "step": 1490, + "valid_targets_mean": 3946.9, + "valid_targets_min": 3217 + }, + { + "epoch": 0.8241455347298787, + "grad_norm": 0.47584580172729574, + "learning_rate": 3.996209311642889e-05, + "loss": 0.1305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06355395168066025, + "step": 1495, + "valid_targets_mean": 2405.0, + "valid_targets_min": 735 + }, + { + "epoch": 0.8269018743109151, + "grad_norm": 0.4897507215458173, + "learning_rate": 3.996038252303184e-05, + "loss": 0.1788, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06079264357686043, + "step": 1500, + "valid_targets_mean": 1502.2, + "valid_targets_min": 603 + }, + { + "epoch": 0.8296582138919515, + "grad_norm": 0.45988665116397465, + "learning_rate": 3.995863421863191e-05, + "loss": 0.1102, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044805798679590225, + "step": 1505, + "valid_targets_mean": 1912.9, + "valid_targets_min": 582 + }, + { + "epoch": 0.8324145534729879, + "grad_norm": 0.5566795932538792, + "learning_rate": 3.995684820653216e-05, + "loss": 0.1327, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0707668885588646, + "step": 1510, + "valid_targets_mean": 2590.9, + "valid_targets_min": 579 + }, + { + "epoch": 0.8351708930540243, + "grad_norm": 0.4066307125273845, + "learning_rate": 3.9955024490106875e-05, + "loss": 0.1173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07581821829080582, + "step": 1515, + "valid_targets_mean": 3709.5, + "valid_targets_min": 3248 + }, + { + "epoch": 0.8379272326350606, + "grad_norm": 0.4968082786303988, + "learning_rate": 3.9953163072801595e-05, + "loss": 0.1795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08675962686538696, + "step": 1520, + "valid_targets_mean": 3212.4, + "valid_targets_min": 994 + }, + { + "epoch": 0.840683572216097, + "grad_norm": 0.569108895654233, + "learning_rate": 3.995126395813309e-05, + "loss": 0.1477, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10220188647508621, + "step": 1525, + "valid_targets_mean": 2010.6, + "valid_targets_min": 692 + }, + { + "epoch": 0.8434399117971334, + "grad_norm": 0.43587122266371875, + "learning_rate": 3.994932714968934e-05, + "loss": 0.1614, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07052162289619446, + "step": 1530, + "valid_targets_mean": 2707.2, + "valid_targets_min": 632 + }, + { + "epoch": 0.8461962513781698, + "grad_norm": 0.44769553117940103, + "learning_rate": 3.994735265112954e-05, + "loss": 0.1419, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05778144672513008, + "step": 1535, + "valid_targets_mean": 1847.8, + "valid_targets_min": 812 + }, + { + "epoch": 0.8489525909592062, + "grad_norm": 0.5133242760553666, + "learning_rate": 3.9945340466184086e-05, + "loss": 0.2056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20955367386341095, + "step": 1540, + "valid_targets_mean": 3245.8, + "valid_targets_min": 1177 + }, + { + "epoch": 0.8517089305402425, + "grad_norm": 0.4556033606340427, + "learning_rate": 3.994329059865461e-05, + "loss": 0.1411, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06725326180458069, + "step": 1545, + "valid_targets_mean": 3384.0, + "valid_targets_min": 2331 + }, + { + "epoch": 0.8544652701212789, + "grad_norm": 0.7105106198763467, + "learning_rate": 3.994120305241389e-05, + "loss": 0.1706, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09072577953338623, + "step": 1550, + "valid_targets_mean": 1604.1, + "valid_targets_min": 684 + }, + { + "epoch": 0.8572216097023153, + "grad_norm": 0.5310134728490914, + "learning_rate": 3.993907783140591e-05, + "loss": 0.1347, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041653137654066086, + "step": 1555, + "valid_targets_mean": 1534.4, + "valid_targets_min": 446 + }, + { + "epoch": 0.8599779492833517, + "grad_norm": 0.4411937356955532, + "learning_rate": 3.993691493964585e-05, + "loss": 0.1408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044283293187618256, + "step": 1560, + "valid_targets_mean": 2990.9, + "valid_targets_min": 655 + }, + { + "epoch": 0.8627342888643881, + "grad_norm": 0.7244288303077998, + "learning_rate": 3.993471438122003e-05, + "loss": 0.1399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06212691217660904, + "step": 1565, + "valid_targets_mean": 1293.9, + "valid_targets_min": 574 + }, + { + "epoch": 0.8654906284454245, + "grad_norm": 0.3777505370792934, + "learning_rate": 3.993247616028596e-05, + "loss": 0.1166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.062912717461586, + "step": 1570, + "valid_targets_mean": 4455.1, + "valid_targets_min": 2872 + }, + { + "epoch": 0.8682469680264608, + "grad_norm": 0.38549202020214024, + "learning_rate": 3.993020028107229e-05, + "loss": 0.1136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06780111789703369, + "step": 1575, + "valid_targets_mean": 3615.2, + "valid_targets_min": 2242 + }, + { + "epoch": 0.8710033076074972, + "grad_norm": 0.4737244470813205, + "learning_rate": 3.9927886747878826e-05, + "loss": 0.1267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057429905980825424, + "step": 1580, + "valid_targets_mean": 2017.4, + "valid_targets_min": 651 + }, + { + "epoch": 0.8737596471885336, + "grad_norm": 0.421152831208703, + "learning_rate": 3.99255355650765e-05, + "loss": 0.1431, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07161194831132889, + "step": 1585, + "valid_targets_mean": 3131.8, + "valid_targets_min": 740 + }, + { + "epoch": 0.87651598676957, + "grad_norm": 0.45035985819448304, + "learning_rate": 3.992314673710739e-05, + "loss": 0.142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08061245828866959, + "step": 1590, + "valid_targets_mean": 3651.6, + "valid_targets_min": 3006 + }, + { + "epoch": 0.8792723263506064, + "grad_norm": 0.34635206951043146, + "learning_rate": 3.992072026848469e-05, + "loss": 0.1294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045042023062705994, + "step": 1595, + "valid_targets_mean": 3182.6, + "valid_targets_min": 2353 + }, + { + "epoch": 0.8820286659316428, + "grad_norm": 0.42878714570304327, + "learning_rate": 3.991825616379271e-05, + "loss": 0.1213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05563933029770851, + "step": 1600, + "valid_targets_mean": 1954.4, + "valid_targets_min": 665 + }, + { + "epoch": 0.8847850055126791, + "grad_norm": 0.4585486894660222, + "learning_rate": 3.9915754427686867e-05, + "loss": 0.1492, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08079366385936737, + "step": 1605, + "valid_targets_mean": 3512.6, + "valid_targets_min": 1776 + }, + { + "epoch": 0.8875413450937155, + "grad_norm": 0.2878302453912057, + "learning_rate": 3.9913215064893664e-05, + "loss": 0.11, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051436375826597214, + "step": 1610, + "valid_targets_mean": 4054.6, + "valid_targets_min": 3675 + }, + { + "epoch": 0.8902976846747519, + "grad_norm": 0.4118312883215123, + "learning_rate": 3.9910638080210714e-05, + "loss": 0.1369, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0586533285677433, + "step": 1615, + "valid_targets_mean": 3238.4, + "valid_targets_min": 591 + }, + { + "epoch": 0.8930540242557883, + "grad_norm": 0.4449998439041385, + "learning_rate": 3.990802347850668e-05, + "loss": 0.1407, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05342760309576988, + "step": 1620, + "valid_targets_mean": 2646.6, + "valid_targets_min": 1011 + }, + { + "epoch": 0.8958103638368247, + "grad_norm": 0.6614818952999044, + "learning_rate": 3.990537126472132e-05, + "loss": 0.1426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.126672625541687, + "step": 1625, + "valid_targets_mean": 1379.1, + "valid_targets_min": 903 + }, + { + "epoch": 0.898566703417861, + "grad_norm": 0.5841064191116052, + "learning_rate": 3.990268144386543e-05, + "loss": 0.1397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06886439025402069, + "step": 1630, + "valid_targets_mean": 2092.5, + "valid_targets_min": 693 + }, + { + "epoch": 0.9013230429988974, + "grad_norm": 0.725524732924324, + "learning_rate": 3.989995402102089e-05, + "loss": 0.1409, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09064992517232895, + "step": 1635, + "valid_targets_mean": 1268.9, + "valid_targets_min": 732 + }, + { + "epoch": 0.9040793825799338, + "grad_norm": 0.4349976326699684, + "learning_rate": 3.989718900134058e-05, + "loss": 0.1313, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04457591474056244, + "step": 1640, + "valid_targets_mean": 2524.1, + "valid_targets_min": 799 + }, + { + "epoch": 0.9068357221609702, + "grad_norm": 0.45704693881149355, + "learning_rate": 3.9894386390048444e-05, + "loss": 0.156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05043622851371765, + "step": 1645, + "valid_targets_mean": 2966.2, + "valid_targets_min": 1012 + }, + { + "epoch": 0.9095920617420066, + "grad_norm": 0.39066073692215286, + "learning_rate": 3.989154619243943e-05, + "loss": 0.1302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06167374551296234, + "step": 1650, + "valid_targets_mean": 3617.0, + "valid_targets_min": 1622 + }, + { + "epoch": 0.912348401323043, + "grad_norm": 0.42731348150660103, + "learning_rate": 3.988866841387949e-05, + "loss": 0.1183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06624037027359009, + "step": 1655, + "valid_targets_mean": 3110.0, + "valid_targets_min": 1001 + }, + { + "epoch": 0.9151047409040793, + "grad_norm": 0.44818291207284705, + "learning_rate": 3.988575305980561e-05, + "loss": 0.1434, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061158135533332825, + "step": 1660, + "valid_targets_mean": 2685.8, + "valid_targets_min": 897 + }, + { + "epoch": 0.9178610804851157, + "grad_norm": 0.33634585529279937, + "learning_rate": 3.988280013572573e-05, + "loss": 0.1396, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05320034176111221, + "step": 1665, + "valid_targets_mean": 3736.1, + "valid_targets_min": 2915 + }, + { + "epoch": 0.9206174200661521, + "grad_norm": 0.5667727814006426, + "learning_rate": 3.9879809647218805e-05, + "loss": 0.1144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08513908088207245, + "step": 1670, + "valid_targets_mean": 2964.5, + "valid_targets_min": 561 + }, + { + "epoch": 0.9233737596471885, + "grad_norm": 0.3657859158105549, + "learning_rate": 3.9876781599934726e-05, + "loss": 0.1263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07076647877693176, + "step": 1675, + "valid_targets_mean": 3624.0, + "valid_targets_min": 2542 + }, + { + "epoch": 0.9261300992282249, + "grad_norm": 0.3174934374959683, + "learning_rate": 3.9873715999594374e-05, + "loss": 0.1133, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05323849618434906, + "step": 1680, + "valid_targets_mean": 4175.0, + "valid_targets_min": 2983 + }, + { + "epoch": 0.9288864388092613, + "grad_norm": 0.4793589291241528, + "learning_rate": 3.987061285198955e-05, + "loss": 0.1054, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04887327179312706, + "step": 1685, + "valid_targets_mean": 2350.6, + "valid_targets_min": 615 + }, + { + "epoch": 0.9316427783902976, + "grad_norm": 0.35671881221119317, + "learning_rate": 3.9867472162983024e-05, + "loss": 0.1344, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045522160828113556, + "step": 1690, + "valid_targets_mean": 3634.9, + "valid_targets_min": 2936 + }, + { + "epoch": 0.934399117971334, + "grad_norm": 0.4171612539579482, + "learning_rate": 3.9864293938508465e-05, + "loss": 0.1237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0902840867638588, + "step": 1695, + "valid_targets_mean": 3372.6, + "valid_targets_min": 1532 + }, + { + "epoch": 0.9371554575523704, + "grad_norm": 0.4015172038177295, + "learning_rate": 3.986107818457048e-05, + "loss": 0.145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039033204317092896, + "step": 1700, + "valid_targets_mean": 2305.8, + "valid_targets_min": 506 + }, + { + "epoch": 0.9399117971334069, + "grad_norm": 0.32690514413636745, + "learning_rate": 3.985782490724455e-05, + "loss": 0.1315, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04189014807343483, + "step": 1705, + "valid_targets_mean": 2746.1, + "valid_targets_min": 633 + }, + { + "epoch": 0.9426681367144433, + "grad_norm": 0.42957220429905285, + "learning_rate": 3.985453411267709e-05, + "loss": 0.1572, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06922756135463715, + "step": 1710, + "valid_targets_mean": 2730.4, + "valid_targets_min": 733 + }, + { + "epoch": 0.9454244762954797, + "grad_norm": 0.34742586926048824, + "learning_rate": 3.985120580708536e-05, + "loss": 0.1144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057322364300489426, + "step": 1715, + "valid_targets_mean": 3070.9, + "valid_targets_min": 489 + }, + { + "epoch": 0.948180815876516, + "grad_norm": 0.5631364781381808, + "learning_rate": 3.98478399967575e-05, + "loss": 0.14, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05616089701652527, + "step": 1720, + "valid_targets_mean": 1371.4, + "valid_targets_min": 693 + }, + { + "epoch": 0.9509371554575524, + "grad_norm": 0.39155749798336426, + "learning_rate": 3.9844436688052526e-05, + "loss": 0.1212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04619358107447624, + "step": 1725, + "valid_targets_mean": 3525.9, + "valid_targets_min": 918 + }, + { + "epoch": 0.9536934950385888, + "grad_norm": 0.6208072643842845, + "learning_rate": 3.9840995887400274e-05, + "loss": 0.1728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07145081460475922, + "step": 1730, + "valid_targets_mean": 1479.6, + "valid_targets_min": 528 + }, + { + "epoch": 0.9564498346196252, + "grad_norm": 0.4764754445544511, + "learning_rate": 3.9837517601301424e-05, + "loss": 0.1558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09725221991539001, + "step": 1735, + "valid_targets_mean": 2083.9, + "valid_targets_min": 572 + }, + { + "epoch": 0.9592061742006616, + "grad_norm": 0.44310643083413837, + "learning_rate": 3.983400183632747e-05, + "loss": 0.127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06867316365242004, + "step": 1740, + "valid_targets_mean": 3714.4, + "valid_targets_min": 2773 + }, + { + "epoch": 0.961962513781698, + "grad_norm": 0.38522253902900344, + "learning_rate": 3.983044859912073e-05, + "loss": 0.1255, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05939926952123642, + "step": 1745, + "valid_targets_mean": 3400.9, + "valid_targets_min": 2395 + }, + { + "epoch": 0.9647188533627343, + "grad_norm": 0.36183084321902653, + "learning_rate": 3.9826857896394306e-05, + "loss": 0.124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04789583012461662, + "step": 1750, + "valid_targets_mean": 2494.9, + "valid_targets_min": 965 + }, + { + "epoch": 0.9674751929437707, + "grad_norm": 0.46159409612657315, + "learning_rate": 3.982322973493209e-05, + "loss": 0.1187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05478442460298538, + "step": 1755, + "valid_targets_mean": 1522.5, + "valid_targets_min": 725 + }, + { + "epoch": 0.9702315325248071, + "grad_norm": 0.7569116640334799, + "learning_rate": 3.9819564121588736e-05, + "loss": 0.1538, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10079911351203918, + "step": 1760, + "valid_targets_mean": 1261.0, + "valid_targets_min": 633 + }, + { + "epoch": 0.9729878721058435, + "grad_norm": 0.3712163235627577, + "learning_rate": 3.981586106328966e-05, + "loss": 0.1127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050364140421152115, + "step": 1765, + "valid_targets_mean": 3052.9, + "valid_targets_min": 889 + }, + { + "epoch": 0.9757442116868799, + "grad_norm": 0.3548575521020409, + "learning_rate": 3.981212056703102e-05, + "loss": 0.1187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05800827592611313, + "step": 1770, + "valid_targets_mean": 3525.2, + "valid_targets_min": 2282 + }, + { + "epoch": 0.9785005512679162, + "grad_norm": 0.438931975315754, + "learning_rate": 3.980834263987971e-05, + "loss": 0.1229, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0647580549120903, + "step": 1775, + "valid_targets_mean": 3317.0, + "valid_targets_min": 981 + }, + { + "epoch": 0.9812568908489526, + "grad_norm": 0.3216018139294105, + "learning_rate": 3.980452728897334e-05, + "loss": 0.1324, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058623552322387695, + "step": 1780, + "valid_targets_mean": 3465.1, + "valid_targets_min": 1071 + }, + { + "epoch": 0.984013230429989, + "grad_norm": 0.3031911671248217, + "learning_rate": 3.9800674521520234e-05, + "loss": 0.1074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051805280148983, + "step": 1785, + "valid_targets_mean": 4006.9, + "valid_targets_min": 757 + }, + { + "epoch": 0.9867695700110254, + "grad_norm": 0.38065169107138574, + "learning_rate": 3.9796784344799385e-05, + "loss": 0.1202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04726122319698334, + "step": 1790, + "valid_targets_mean": 1548.0, + "valid_targets_min": 504 + }, + { + "epoch": 0.9895259095920618, + "grad_norm": 0.533943349001374, + "learning_rate": 3.979285676616048e-05, + "loss": 0.1384, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11262380331754684, + "step": 1795, + "valid_targets_mean": 2691.9, + "valid_targets_min": 775 + }, + { + "epoch": 0.9922822491730982, + "grad_norm": 0.3941601661164578, + "learning_rate": 3.978889179302385e-05, + "loss": 0.2087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058255575597286224, + "step": 1800, + "valid_targets_mean": 3880.9, + "valid_targets_min": 2338 + }, + { + "epoch": 0.9950385887541345, + "grad_norm": 0.44183813622656276, + "learning_rate": 3.9784889432880513e-05, + "loss": 0.1203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07612086087465286, + "step": 1805, + "valid_targets_mean": 3656.0, + "valid_targets_min": 2463 + }, + { + "epoch": 0.9977949283351709, + "grad_norm": 0.3851099846546952, + "learning_rate": 3.9780849693292094e-05, + "loss": 0.1193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05513102188706398, + "step": 1810, + "valid_targets_mean": 3262.0, + "valid_targets_min": 2237 + }, + { + "epoch": 1.0005512679162072, + "grad_norm": 0.5253719737295299, + "learning_rate": 3.977677258189082e-05, + "loss": 0.1924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1398051679134369, + "step": 1815, + "valid_targets_mean": 8186.1, + "valid_targets_min": 6131 + }, + { + "epoch": 1.0033076074972436, + "grad_norm": 0.4243427479818193, + "learning_rate": 3.977265810637957e-05, + "loss": 0.2826, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1418125033378601, + "step": 1820, + "valid_targets_mean": 8564.9, + "valid_targets_min": 5974 + }, + { + "epoch": 1.00606394707828, + "grad_norm": 0.5713760549516356, + "learning_rate": 3.9768506274531795e-05, + "loss": 0.2761, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1232951432466507, + "step": 1825, + "valid_targets_mean": 3567.1, + "valid_targets_min": 287 + }, + { + "epoch": 1.0088202866593163, + "grad_norm": 0.3978748828028442, + "learning_rate": 3.97643170941915e-05, + "loss": 0.2648, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13080854713916779, + "step": 1830, + "valid_targets_mean": 6991.4, + "valid_targets_min": 4953 + }, + { + "epoch": 1.0115766262403527, + "grad_norm": 0.43207704765130583, + "learning_rate": 3.9760090573273286e-05, + "loss": 0.2426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12646427750587463, + "step": 1835, + "valid_targets_mean": 6929.9, + "valid_targets_min": 1449 + }, + { + "epoch": 1.014332965821389, + "grad_norm": 0.390558427850409, + "learning_rate": 3.975582671976229e-05, + "loss": 0.2492, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11817316710948944, + "step": 1840, + "valid_targets_mean": 7335.5, + "valid_targets_min": 4961 + }, + { + "epoch": 1.0170893054024255, + "grad_norm": 0.4277711641408585, + "learning_rate": 3.9751525541714164e-05, + "loss": 0.2653, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15716847777366638, + "step": 1845, + "valid_targets_mean": 7028.9, + "valid_targets_min": 5952 + }, + { + "epoch": 1.0198456449834619, + "grad_norm": 0.38360211997332955, + "learning_rate": 3.974718704725511e-05, + "loss": 0.2613, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1409161388874054, + "step": 1850, + "valid_targets_mean": 7634.2, + "valid_targets_min": 5143 + }, + { + "epoch": 1.0226019845644982, + "grad_norm": 0.3907357641216315, + "learning_rate": 3.9742811244581806e-05, + "loss": 0.2496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12707455456256866, + "step": 1855, + "valid_targets_mean": 7079.8, + "valid_targets_min": 5316 + }, + { + "epoch": 1.0253583241455346, + "grad_norm": 0.503067349336614, + "learning_rate": 3.973839814196142e-05, + "loss": 0.2467, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11652622371912003, + "step": 1860, + "valid_targets_mean": 4218.1, + "valid_targets_min": 151 + }, + { + "epoch": 1.028114663726571, + "grad_norm": 0.37549183347504983, + "learning_rate": 3.973394774773159e-05, + "loss": 0.2412, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12704139947891235, + "step": 1865, + "valid_targets_mean": 7171.6, + "valid_targets_min": 4244 + }, + { + "epoch": 1.0308710033076074, + "grad_norm": 0.3726249942841516, + "learning_rate": 3.972946007030043e-05, + "loss": 0.2433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11867683380842209, + "step": 1870, + "valid_targets_mean": 7214.5, + "valid_targets_min": 5799 + }, + { + "epoch": 1.0336273428886438, + "grad_norm": 0.4286770118825725, + "learning_rate": 3.972493511814646e-05, + "loss": 0.2395, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11824707686901093, + "step": 1875, + "valid_targets_mean": 6923.1, + "valid_targets_min": 3844 + }, + { + "epoch": 1.0363836824696802, + "grad_norm": 0.38263404295353465, + "learning_rate": 3.972037289981865e-05, + "loss": 0.237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12816131114959717, + "step": 1880, + "valid_targets_mean": 6961.0, + "valid_targets_min": 5664 + }, + { + "epoch": 1.0391400220507165, + "grad_norm": 0.4109650145403731, + "learning_rate": 3.971577342393635e-05, + "loss": 0.2295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12724187970161438, + "step": 1885, + "valid_targets_mean": 6933.6, + "valid_targets_min": 4915 + }, + { + "epoch": 1.041896361631753, + "grad_norm": 0.37457751520573856, + "learning_rate": 3.971113669918933e-05, + "loss": 0.2349, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11184582114219666, + "step": 1890, + "valid_targets_mean": 6407.2, + "valid_targets_min": 4966 + }, + { + "epoch": 1.0446527012127893, + "grad_norm": 0.3818615223468083, + "learning_rate": 3.9706462734337714e-05, + "loss": 0.2371, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11798550933599472, + "step": 1895, + "valid_targets_mean": 7489.0, + "valid_targets_min": 5451 + }, + { + "epoch": 1.0474090407938257, + "grad_norm": 0.37325279832695785, + "learning_rate": 3.970175153821199e-05, + "loss": 0.2288, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10837207734584808, + "step": 1900, + "valid_targets_mean": 7392.1, + "valid_targets_min": 5195 + }, + { + "epoch": 1.0501653803748623, + "grad_norm": 0.40704113963058447, + "learning_rate": 3.9697003119712994e-05, + "loss": 0.2488, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1524529904127121, + "step": 1905, + "valid_targets_mean": 7050.0, + "valid_targets_min": 3733 + }, + { + "epoch": 1.0529217199558987, + "grad_norm": 0.45258933611038, + "learning_rate": 3.969221748781186e-05, + "loss": 0.2605, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13360163569450378, + "step": 1910, + "valid_targets_mean": 6038.4, + "valid_targets_min": 5143 + }, + { + "epoch": 1.055678059536935, + "grad_norm": 0.38249642566511843, + "learning_rate": 3.9687394651550055e-05, + "loss": 0.219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11315138638019562, + "step": 1915, + "valid_targets_mean": 7356.9, + "valid_targets_min": 5765 + }, + { + "epoch": 1.0584343991179714, + "grad_norm": 0.3609942898084209, + "learning_rate": 3.968253462003934e-05, + "loss": 0.2228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10660286247730255, + "step": 1920, + "valid_targets_mean": 7668.2, + "valid_targets_min": 4640 + }, + { + "epoch": 1.0611907386990078, + "grad_norm": 0.3963853311905443, + "learning_rate": 3.967763740246172e-05, + "loss": 0.2293, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12513354420661926, + "step": 1925, + "valid_targets_mean": 7758.0, + "valid_targets_min": 5483 + }, + { + "epoch": 1.0639470782800442, + "grad_norm": 0.40502854294111473, + "learning_rate": 3.967270300806948e-05, + "loss": 0.2413, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.117188461124897, + "step": 1930, + "valid_targets_mean": 7311.8, + "valid_targets_min": 6424 + }, + { + "epoch": 1.0667034178610806, + "grad_norm": 0.38258033688936405, + "learning_rate": 3.9667731446185126e-05, + "loss": 0.232, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10352829098701477, + "step": 1935, + "valid_targets_mean": 6788.4, + "valid_targets_min": 4371 + }, + { + "epoch": 1.069459757442117, + "grad_norm": 0.43996408006926535, + "learning_rate": 3.966272272620141e-05, + "loss": 0.2311, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10274834930896759, + "step": 1940, + "valid_targets_mean": 7830.8, + "valid_targets_min": 5371 + }, + { + "epoch": 1.0722160970231533, + "grad_norm": 0.34184844127688635, + "learning_rate": 3.965767685758126e-05, + "loss": 0.2023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09315471351146698, + "step": 1945, + "valid_targets_mean": 8286.8, + "valid_targets_min": 4818 + }, + { + "epoch": 1.0749724366041897, + "grad_norm": 0.3765936208330506, + "learning_rate": 3.965259384985779e-05, + "loss": 0.2046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08744172751903534, + "step": 1950, + "valid_targets_mean": 7339.1, + "valid_targets_min": 5995 + }, + { + "epoch": 1.0777287761852261, + "grad_norm": 0.40929348356782497, + "learning_rate": 3.96474737126343e-05, + "loss": 0.214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11613349616527557, + "step": 1955, + "valid_targets_mean": 6405.4, + "valid_targets_min": 4206 + }, + { + "epoch": 1.0804851157662625, + "grad_norm": 1.699550098485449, + "learning_rate": 3.964231645558422e-05, + "loss": 0.237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10056192427873611, + "step": 1960, + "valid_targets_mean": 246.4, + "valid_targets_min": 136 + }, + { + "epoch": 1.0832414553472989, + "grad_norm": 0.4050918665505418, + "learning_rate": 3.9637122088451125e-05, + "loss": 0.2412, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11114350706338882, + "step": 1965, + "valid_targets_mean": 6778.4, + "valid_targets_min": 5133 + }, + { + "epoch": 1.0859977949283353, + "grad_norm": 0.4265782338317279, + "learning_rate": 3.963189062104869e-05, + "loss": 0.2454, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12836581468582153, + "step": 1970, + "valid_targets_mean": 6963.0, + "valid_targets_min": 5072 + }, + { + "epoch": 1.0887541345093716, + "grad_norm": 0.3878132159900477, + "learning_rate": 3.962662206326069e-05, + "loss": 0.2438, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13163337111473083, + "step": 1975, + "valid_targets_mean": 6662.5, + "valid_targets_min": 5036 + }, + { + "epoch": 1.091510474090408, + "grad_norm": 0.42456212394648, + "learning_rate": 3.962131642504097e-05, + "loss": 0.2372, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14148050546646118, + "step": 1980, + "valid_targets_mean": 7389.0, + "valid_targets_min": 5546 + }, + { + "epoch": 1.0942668136714444, + "grad_norm": 0.3787115022020608, + "learning_rate": 3.9615973716413434e-05, + "loss": 0.2262, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11571347713470459, + "step": 1985, + "valid_targets_mean": 7507.0, + "valid_targets_min": 5051 + }, + { + "epoch": 1.0970231532524808, + "grad_norm": 0.3587565393870691, + "learning_rate": 3.961059394747202e-05, + "loss": 0.2351, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10655775666236877, + "step": 1990, + "valid_targets_mean": 7192.6, + "valid_targets_min": 5809 + }, + { + "epoch": 1.0997794928335172, + "grad_norm": 0.3763775185330977, + "learning_rate": 3.9605177128380686e-05, + "loss": 0.2285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11339762806892395, + "step": 1995, + "valid_targets_mean": 7255.8, + "valid_targets_min": 5119 + }, + { + "epoch": 1.1025358324145536, + "grad_norm": 0.39553415900581035, + "learning_rate": 3.9599723269373393e-05, + "loss": 0.2391, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11103244870901108, + "step": 2000, + "valid_targets_mean": 6710.9, + "valid_targets_min": 4029 + }, + { + "epoch": 1.10529217199559, + "grad_norm": 0.45471435552416234, + "learning_rate": 3.959423238075407e-05, + "loss": 0.2296, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12882176041603088, + "step": 2005, + "valid_targets_mean": 6299.6, + "valid_targets_min": 5464 + }, + { + "epoch": 1.1080485115766263, + "grad_norm": 0.42526036861014666, + "learning_rate": 3.958870447289662e-05, + "loss": 0.2343, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11438222974538803, + "step": 2010, + "valid_targets_mean": 6239.8, + "valid_targets_min": 4663 + }, + { + "epoch": 1.1108048511576627, + "grad_norm": 0.42567878407948523, + "learning_rate": 3.958313955624487e-05, + "loss": 0.2353, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12137986719608307, + "step": 2015, + "valid_targets_mean": 5759.1, + "valid_targets_min": 4989 + }, + { + "epoch": 1.113561190738699, + "grad_norm": 0.46986674473654827, + "learning_rate": 3.957753764131258e-05, + "loss": 0.2262, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09319612383842468, + "step": 2020, + "valid_targets_mean": 3368.0, + "valid_targets_min": 2264 + }, + { + "epoch": 1.1163175303197355, + "grad_norm": 0.4434810204982593, + "learning_rate": 3.9571898738683394e-05, + "loss": 0.2278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11361192911863327, + "step": 2025, + "valid_targets_mean": 6208.5, + "valid_targets_min": 5061 + }, + { + "epoch": 1.1190738699007718, + "grad_norm": 0.40004861980867673, + "learning_rate": 3.9566222859010875e-05, + "loss": 0.2274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1181272640824318, + "step": 2030, + "valid_targets_mean": 6299.4, + "valid_targets_min": 4566 + }, + { + "epoch": 1.1218302094818082, + "grad_norm": 0.3950043951079505, + "learning_rate": 3.95605100130184e-05, + "loss": 0.2264, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12107467651367188, + "step": 2035, + "valid_targets_mean": 7560.9, + "valid_targets_min": 4931 + }, + { + "epoch": 1.1245865490628446, + "grad_norm": 0.40395212525538315, + "learning_rate": 3.9554760211499214e-05, + "loss": 0.2271, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09796209633350372, + "step": 2040, + "valid_targets_mean": 5546.4, + "valid_targets_min": 4426 + }, + { + "epoch": 1.127342888643881, + "grad_norm": 0.45608132404198176, + "learning_rate": 3.9548973465316374e-05, + "loss": 0.2257, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10861669480800629, + "step": 2045, + "valid_targets_mean": 5400.6, + "valid_targets_min": 4401 + }, + { + "epoch": 1.1300992282249174, + "grad_norm": 0.39779025491490827, + "learning_rate": 3.9543149785402733e-05, + "loss": 0.2297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1172780692577362, + "step": 2050, + "valid_targets_mean": 6330.9, + "valid_targets_min": 5178 + }, + { + "epoch": 1.1328555678059538, + "grad_norm": 0.4436564243441477, + "learning_rate": 3.9537289182760926e-05, + "loss": 0.2303, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1058489978313446, + "step": 2055, + "valid_targets_mean": 5793.9, + "valid_targets_min": 5097 + }, + { + "epoch": 1.1356119073869901, + "grad_norm": 0.836519897672139, + "learning_rate": 3.953139166846335e-05, + "loss": 0.2044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16175433993339539, + "step": 2060, + "valid_targets_mean": 1627.8, + "valid_targets_min": 136 + }, + { + "epoch": 1.1383682469680265, + "grad_norm": 0.38320762563795757, + "learning_rate": 3.952545725365215e-05, + "loss": 0.2255, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11359821259975433, + "step": 2065, + "valid_targets_mean": 6442.6, + "valid_targets_min": 4719 + }, + { + "epoch": 1.141124586549063, + "grad_norm": 0.39206418594165965, + "learning_rate": 3.951948594953914e-05, + "loss": 0.2218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12199763208627701, + "step": 2070, + "valid_targets_mean": 7150.1, + "valid_targets_min": 5326 + }, + { + "epoch": 1.1438809261300993, + "grad_norm": 0.37270153920091903, + "learning_rate": 3.951347776740588e-05, + "loss": 0.2286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11454448103904724, + "step": 2075, + "valid_targets_mean": 7101.2, + "valid_targets_min": 4056 + }, + { + "epoch": 1.1466372657111357, + "grad_norm": 0.39336139751094895, + "learning_rate": 3.9507432718603584e-05, + "loss": 0.2204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11072040349245071, + "step": 2080, + "valid_targets_mean": 6992.9, + "valid_targets_min": 4903 + }, + { + "epoch": 1.149393605292172, + "grad_norm": 0.4026382151474578, + "learning_rate": 3.9501350814553114e-05, + "loss": 0.2149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11492349207401276, + "step": 2085, + "valid_targets_mean": 5869.9, + "valid_targets_min": 4747 + }, + { + "epoch": 1.1521499448732084, + "grad_norm": 0.39802797850832144, + "learning_rate": 3.9495232066744965e-05, + "loss": 0.2253, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11544560641050339, + "step": 2090, + "valid_targets_mean": 6258.1, + "valid_targets_min": 4588 + }, + { + "epoch": 1.1549062844542448, + "grad_norm": 1.1432856855045128, + "learning_rate": 3.9489076486739245e-05, + "loss": 0.2466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0995759591460228, + "step": 2095, + "valid_targets_mean": 924.5, + "valid_targets_min": 533 + }, + { + "epoch": 1.1576626240352812, + "grad_norm": 0.8984213162431898, + "learning_rate": 3.948288408616565e-05, + "loss": 0.2454, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10126940160989761, + "step": 2100, + "valid_targets_mean": 1198.2, + "valid_targets_min": 688 + }, + { + "epoch": 1.1604189636163176, + "grad_norm": 0.8805805500907204, + "learning_rate": 3.947665487672342e-05, + "loss": 0.2367, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11180521547794342, + "step": 2105, + "valid_targets_mean": 1214.4, + "valid_targets_min": 655 + }, + { + "epoch": 1.163175303197354, + "grad_norm": 0.9879810382935381, + "learning_rate": 3.9470388870181376e-05, + "loss": 0.231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12215855717658997, + "step": 2110, + "valid_targets_mean": 1183.0, + "valid_targets_min": 579 + }, + { + "epoch": 1.1659316427783903, + "grad_norm": 0.7600754133546123, + "learning_rate": 3.9464086078377824e-05, + "loss": 0.2313, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10856646299362183, + "step": 2115, + "valid_targets_mean": 1302.1, + "valid_targets_min": 494 + }, + { + "epoch": 1.1686879823594267, + "grad_norm": 0.767723571405227, + "learning_rate": 3.945774651322058e-05, + "loss": 0.2239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11804922670125961, + "step": 2120, + "valid_targets_mean": 1632.4, + "valid_targets_min": 723 + }, + { + "epoch": 1.171444321940463, + "grad_norm": 0.8255616132396073, + "learning_rate": 3.945137018668695e-05, + "loss": 0.2333, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08698814362287521, + "step": 2125, + "valid_targets_mean": 1045.1, + "valid_targets_min": 631 + }, + { + "epoch": 1.1742006615214995, + "grad_norm": 0.9610015777759753, + "learning_rate": 3.9444957110823665e-05, + "loss": 0.22, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13884054124355316, + "step": 2130, + "valid_targets_mean": 1272.4, + "valid_targets_min": 711 + }, + { + "epoch": 1.1769570011025359, + "grad_norm": 0.790140243624563, + "learning_rate": 3.943850729774692e-05, + "loss": 0.2431, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10402952879667282, + "step": 2135, + "valid_targets_mean": 1190.1, + "valid_targets_min": 649 + }, + { + "epoch": 1.1797133406835723, + "grad_norm": 0.7959918216104902, + "learning_rate": 3.943202075964229e-05, + "loss": 0.2308, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12850704789161682, + "step": 2140, + "valid_targets_mean": 1823.4, + "valid_targets_min": 788 + }, + { + "epoch": 1.1824696802646086, + "grad_norm": 0.7850835093122766, + "learning_rate": 3.942549750876473e-05, + "loss": 0.2215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11539431661367416, + "step": 2145, + "valid_targets_mean": 1319.8, + "valid_targets_min": 552 + }, + { + "epoch": 1.185226019845645, + "grad_norm": 0.7476705570939407, + "learning_rate": 3.941893755743859e-05, + "loss": 0.2271, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12577146291732788, + "step": 2150, + "valid_targets_mean": 1626.4, + "valid_targets_min": 1005 + }, + { + "epoch": 1.1879823594266814, + "grad_norm": 0.7598217319901422, + "learning_rate": 3.9412340918057525e-05, + "loss": 0.2244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09213559329509735, + "step": 2155, + "valid_targets_mean": 1193.9, + "valid_targets_min": 495 + }, + { + "epoch": 1.1907386990077178, + "grad_norm": 0.7699286552289648, + "learning_rate": 3.9405707603084525e-05, + "loss": 0.2038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09145444631576538, + "step": 2160, + "valid_targets_mean": 1010.6, + "valid_targets_min": 538 + }, + { + "epoch": 1.1934950385887542, + "grad_norm": 0.9019704787995697, + "learning_rate": 3.939903762505186e-05, + "loss": 0.2081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1084647923707962, + "step": 2165, + "valid_targets_mean": 1310.4, + "valid_targets_min": 639 + }, + { + "epoch": 1.1962513781697905, + "grad_norm": 0.9981129660583756, + "learning_rate": 3.9392330996561064e-05, + "loss": 0.2137, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07070482522249222, + "step": 2170, + "valid_targets_mean": 914.5, + "valid_targets_min": 540 + }, + { + "epoch": 1.199007717750827, + "grad_norm": 0.9788738166330087, + "learning_rate": 3.9385587730282936e-05, + "loss": 0.2286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10725310444831848, + "step": 2175, + "valid_targets_mean": 1156.4, + "valid_targets_min": 629 + }, + { + "epoch": 1.2017640573318633, + "grad_norm": 0.8805947566768758, + "learning_rate": 3.937880783895746e-05, + "loss": 0.2306, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13779644668102264, + "step": 2180, + "valid_targets_mean": 1455.2, + "valid_targets_min": 506 + }, + { + "epoch": 1.2045203969128997, + "grad_norm": 0.8685192204773639, + "learning_rate": 3.9371991335393846e-05, + "loss": 0.2117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09417331963777542, + "step": 2185, + "valid_targets_mean": 1249.9, + "valid_targets_min": 553 + }, + { + "epoch": 1.207276736493936, + "grad_norm": 0.7512537682111782, + "learning_rate": 3.936513823247047e-05, + "loss": 0.2233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09678730368614197, + "step": 2190, + "valid_targets_mean": 1130.6, + "valid_targets_min": 488 + }, + { + "epoch": 1.2100330760749725, + "grad_norm": 0.9049087783176254, + "learning_rate": 3.935824854313483e-05, + "loss": 0.2237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13375982642173767, + "step": 2195, + "valid_targets_mean": 1354.8, + "valid_targets_min": 620 + }, + { + "epoch": 1.2127894156560088, + "grad_norm": 0.8646687027723606, + "learning_rate": 3.935132228040357e-05, + "loss": 0.2266, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12267838418483734, + "step": 2200, + "valid_targets_mean": 1189.6, + "valid_targets_min": 690 + }, + { + "epoch": 1.2155457552370452, + "grad_norm": 1.089085094274761, + "learning_rate": 3.934435945736244e-05, + "loss": 0.2183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05583695322275162, + "step": 2205, + "valid_targets_mean": 658.8, + "valid_targets_min": 460 + }, + { + "epoch": 1.2183020948180816, + "grad_norm": 0.9063009384942541, + "learning_rate": 3.933736008716624e-05, + "loss": 0.2149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11861489713191986, + "step": 2210, + "valid_targets_mean": 1305.6, + "valid_targets_min": 499 + }, + { + "epoch": 1.221058434399118, + "grad_norm": 0.7896541478043261, + "learning_rate": 3.933032418303883e-05, + "loss": 0.2147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12515732645988464, + "step": 2215, + "valid_targets_mean": 1593.8, + "valid_targets_min": 561 + }, + { + "epoch": 1.2238147739801544, + "grad_norm": 0.7722440193231476, + "learning_rate": 3.932325175827309e-05, + "loss": 0.218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1277981996536255, + "step": 2220, + "valid_targets_mean": 1541.1, + "valid_targets_min": 1057 + }, + { + "epoch": 1.2265711135611908, + "grad_norm": 0.8166396842709093, + "learning_rate": 3.9316142826230905e-05, + "loss": 0.2197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13855129480361938, + "step": 2225, + "valid_targets_mean": 1733.4, + "valid_targets_min": 1103 + }, + { + "epoch": 1.2293274531422271, + "grad_norm": 1.3497002251270316, + "learning_rate": 3.9308997400343114e-05, + "loss": 0.217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0952853336930275, + "step": 2230, + "valid_targets_mean": 1164.4, + "valid_targets_min": 635 + }, + { + "epoch": 1.2320837927232635, + "grad_norm": 0.8034102232774566, + "learning_rate": 3.930181549410954e-05, + "loss": 0.2059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15352743864059448, + "step": 2235, + "valid_targets_mean": 1747.9, + "valid_targets_min": 786 + }, + { + "epoch": 1.2348401323043, + "grad_norm": 0.7854157617276796, + "learning_rate": 3.9294597121098884e-05, + "loss": 0.2098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11042620241641998, + "step": 2240, + "valid_targets_mean": 1471.6, + "valid_targets_min": 697 + }, + { + "epoch": 1.2375964718853363, + "grad_norm": 0.8808749783557508, + "learning_rate": 3.928734229494878e-05, + "loss": 0.2154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11738111823797226, + "step": 2245, + "valid_targets_mean": 1413.9, + "valid_targets_min": 612 + }, + { + "epoch": 1.2403528114663727, + "grad_norm": 0.8453948819039694, + "learning_rate": 3.928005102936571e-05, + "loss": 0.2202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10927804559469223, + "step": 2250, + "valid_targets_mean": 1360.5, + "valid_targets_min": 1015 + }, + { + "epoch": 1.243109151047409, + "grad_norm": 0.8438828025388841, + "learning_rate": 3.9272723338125e-05, + "loss": 0.2121, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07762035727500916, + "step": 2255, + "valid_targets_mean": 810.5, + "valid_targets_min": 521 + }, + { + "epoch": 1.2458654906284454, + "grad_norm": 0.8788268799877215, + "learning_rate": 3.926535923507081e-05, + "loss": 0.2055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12023581564426422, + "step": 2260, + "valid_targets_mean": 1617.4, + "valid_targets_min": 815 + }, + { + "epoch": 1.2486218302094818, + "grad_norm": 1.0867919496543434, + "learning_rate": 3.92579587341161e-05, + "loss": 0.1972, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10071377456188202, + "step": 2265, + "valid_targets_mean": 1233.5, + "valid_targets_min": 605 + }, + { + "epoch": 1.2513781697905182, + "grad_norm": 0.8681502251088379, + "learning_rate": 3.9250521849242555e-05, + "loss": 0.2163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0718332901597023, + "step": 2270, + "valid_targets_mean": 799.8, + "valid_targets_min": 478 + }, + { + "epoch": 1.2541345093715546, + "grad_norm": 0.7321358752049405, + "learning_rate": 3.9243048594500647e-05, + "loss": 0.2089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09201675653457642, + "step": 2275, + "valid_targets_mean": 1218.1, + "valid_targets_min": 522 + }, + { + "epoch": 1.256890848952591, + "grad_norm": 0.8056814926957335, + "learning_rate": 3.923553898400953e-05, + "loss": 0.2005, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09378452599048615, + "step": 2280, + "valid_targets_mean": 1148.9, + "valid_targets_min": 539 + }, + { + "epoch": 1.2596471885336273, + "grad_norm": 1.1705289444349416, + "learning_rate": 3.922799303195707e-05, + "loss": 0.215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12225021421909332, + "step": 2285, + "valid_targets_mean": 1526.0, + "valid_targets_min": 539 + }, + { + "epoch": 1.2624035281146637, + "grad_norm": 0.8669856496246496, + "learning_rate": 3.922041075259975e-05, + "loss": 0.2193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12422262877225876, + "step": 2290, + "valid_targets_mean": 1506.6, + "valid_targets_min": 1024 + }, + { + "epoch": 1.2651598676957, + "grad_norm": 1.0643861701475001, + "learning_rate": 3.921279216026275e-05, + "loss": 0.2016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13884949684143066, + "step": 2295, + "valid_targets_mean": 1525.4, + "valid_targets_min": 579 + }, + { + "epoch": 1.2679162072767365, + "grad_norm": 0.8343186573601244, + "learning_rate": 3.920513726933979e-05, + "loss": 0.2171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12819573283195496, + "step": 2300, + "valid_targets_mean": 1283.8, + "valid_targets_min": 693 + }, + { + "epoch": 1.2706725468577729, + "grad_norm": 4.205240446116326, + "learning_rate": 3.919744609429321e-05, + "loss": 0.2156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10665176808834076, + "step": 2305, + "valid_targets_mean": 1233.0, + "valid_targets_min": 944 + }, + { + "epoch": 1.2734288864388092, + "grad_norm": 0.7972077466789925, + "learning_rate": 3.918971864965389e-05, + "loss": 0.2254, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11743055284023285, + "step": 2310, + "valid_targets_mean": 1430.8, + "valid_targets_min": 471 + }, + { + "epoch": 1.2761852260198456, + "grad_norm": 0.8076287218221311, + "learning_rate": 3.9181954950021236e-05, + "loss": 0.2046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12675753235816956, + "step": 2315, + "valid_targets_mean": 1406.9, + "valid_targets_min": 626 + }, + { + "epoch": 1.278941565600882, + "grad_norm": 0.8035178915927854, + "learning_rate": 3.917415501006315e-05, + "loss": 0.212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10982207953929901, + "step": 2320, + "valid_targets_mean": 1598.9, + "valid_targets_min": 793 + }, + { + "epoch": 1.2816979051819184, + "grad_norm": 0.8442204304313352, + "learning_rate": 3.9166318844516e-05, + "loss": 0.2073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10271835327148438, + "step": 2325, + "valid_targets_mean": 1102.5, + "valid_targets_min": 546 + }, + { + "epoch": 1.2844542447629548, + "grad_norm": 0.8049118959844419, + "learning_rate": 3.915844646818459e-05, + "loss": 0.1982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0786391943693161, + "step": 2330, + "valid_targets_mean": 960.6, + "valid_targets_min": 463 + }, + { + "epoch": 1.2872105843439912, + "grad_norm": 0.783087914797136, + "learning_rate": 3.915053789594214e-05, + "loss": 0.2052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10732211172580719, + "step": 2335, + "valid_targets_mean": 1476.2, + "valid_targets_min": 502 + }, + { + "epoch": 1.2899669239250275, + "grad_norm": 0.8058135921426168, + "learning_rate": 3.914259314273027e-05, + "loss": 0.2065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11726048588752747, + "step": 2340, + "valid_targets_mean": 1639.6, + "valid_targets_min": 791 + }, + { + "epoch": 1.292723263506064, + "grad_norm": 0.805682137928642, + "learning_rate": 3.913461222355893e-05, + "loss": 0.2062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07545138895511627, + "step": 2345, + "valid_targets_mean": 1105.2, + "valid_targets_min": 650 + }, + { + "epoch": 1.2954796030871003, + "grad_norm": 0.8412493145567457, + "learning_rate": 3.9126595153506424e-05, + "loss": 0.2109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0779571682214737, + "step": 2350, + "valid_targets_mean": 1122.5, + "valid_targets_min": 783 + }, + { + "epoch": 1.2982359426681367, + "grad_norm": 0.8166146514381003, + "learning_rate": 3.9118541947719334e-05, + "loss": 0.2026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08342064172029495, + "step": 2355, + "valid_targets_mean": 1240.1, + "valid_targets_min": 594 + }, + { + "epoch": 1.300992282249173, + "grad_norm": 0.8149687616138347, + "learning_rate": 3.911045262141252e-05, + "loss": 0.201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10064040869474411, + "step": 2360, + "valid_targets_mean": 1214.1, + "valid_targets_min": 599 + }, + { + "epoch": 1.3037486218302095, + "grad_norm": 0.8901006451119075, + "learning_rate": 3.910232718986909e-05, + "loss": 0.2095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12973937392234802, + "step": 2365, + "valid_targets_mean": 1340.0, + "valid_targets_min": 708 + }, + { + "epoch": 1.3065049614112458, + "grad_norm": 0.8186950344164013, + "learning_rate": 3.909416566844036e-05, + "loss": 0.2013, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07486046105623245, + "step": 2370, + "valid_targets_mean": 949.5, + "valid_targets_min": 626 + }, + { + "epoch": 1.3092613009922822, + "grad_norm": 0.776345771859385, + "learning_rate": 3.908596807254585e-05, + "loss": 0.2101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11124133318662643, + "step": 2375, + "valid_targets_mean": 1736.5, + "valid_targets_min": 940 + }, + { + "epoch": 1.3120176405733186, + "grad_norm": 1.0825860662689575, + "learning_rate": 3.9077734417673186e-05, + "loss": 0.189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09750264883041382, + "step": 2380, + "valid_targets_mean": 1206.4, + "valid_targets_min": 641 + }, + { + "epoch": 1.314773980154355, + "grad_norm": 1.5717540424565932, + "learning_rate": 3.906946471937817e-05, + "loss": 0.2055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12011203169822693, + "step": 2385, + "valid_targets_mean": 1205.4, + "valid_targets_min": 492 + }, + { + "epoch": 1.3175303197353914, + "grad_norm": 0.9150967064727711, + "learning_rate": 3.906115899328468e-05, + "loss": 0.2086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06157895550131798, + "step": 2390, + "valid_targets_mean": 1165.6, + "valid_targets_min": 592 + }, + { + "epoch": 1.3202866593164277, + "grad_norm": 0.8792352003377548, + "learning_rate": 3.905281725508466e-05, + "loss": 0.193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07310234755277634, + "step": 2395, + "valid_targets_mean": 831.5, + "valid_targets_min": 533 + }, + { + "epoch": 1.3230429988974641, + "grad_norm": 0.8495793013838969, + "learning_rate": 3.90444395205381e-05, + "loss": 0.2054, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11875495314598083, + "step": 2400, + "valid_targets_mean": 1535.8, + "valid_targets_min": 1072 + }, + { + "epoch": 1.3257993384785005, + "grad_norm": 0.87835167536269, + "learning_rate": 3.9036025805472976e-05, + "loss": 0.2032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0874565988779068, + "step": 2405, + "valid_targets_mean": 1162.1, + "valid_targets_min": 576 + }, + { + "epoch": 1.328555678059537, + "grad_norm": 0.9044763894799498, + "learning_rate": 3.902757612578528e-05, + "loss": 0.2112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09972873330116272, + "step": 2410, + "valid_targets_mean": 1351.5, + "valid_targets_min": 505 + }, + { + "epoch": 1.3313120176405733, + "grad_norm": 1.067365933602101, + "learning_rate": 3.9019090497438916e-05, + "loss": 0.207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1004686951637268, + "step": 2415, + "valid_targets_mean": 1197.6, + "valid_targets_min": 747 + }, + { + "epoch": 1.3340683572216097, + "grad_norm": 0.8937402001301628, + "learning_rate": 3.901056893646572e-05, + "loss": 0.2003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10233119130134583, + "step": 2420, + "valid_targets_mean": 1411.5, + "valid_targets_min": 1107 + }, + { + "epoch": 1.336824696802646, + "grad_norm": 1.0531553248146455, + "learning_rate": 3.900201145896541e-05, + "loss": 0.2132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09617359936237335, + "step": 2425, + "valid_targets_mean": 1218.2, + "valid_targets_min": 579 + }, + { + "epoch": 1.3395810363836824, + "grad_norm": 0.8657147083621377, + "learning_rate": 3.899341808110558e-05, + "loss": 0.2078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08935582637786865, + "step": 2430, + "valid_targets_mean": 1226.9, + "valid_targets_min": 535 + }, + { + "epoch": 1.3423373759647188, + "grad_norm": 0.8217852052715539, + "learning_rate": 3.898478881912162e-05, + "loss": 0.2112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.125512033700943, + "step": 2435, + "valid_targets_mean": 1649.1, + "valid_targets_min": 610 + }, + { + "epoch": 1.3450937155457552, + "grad_norm": 0.9297452385924422, + "learning_rate": 3.897612368931674e-05, + "loss": 0.2163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09725108742713928, + "step": 2440, + "valid_targets_mean": 1128.8, + "valid_targets_min": 720 + }, + { + "epoch": 1.3478500551267916, + "grad_norm": 0.8924688794252034, + "learning_rate": 3.896742270806189e-05, + "loss": 0.2114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11214599013328552, + "step": 2445, + "valid_targets_mean": 1602.5, + "valid_targets_min": 1013 + }, + { + "epoch": 1.350606394707828, + "grad_norm": 1.0714094293202363, + "learning_rate": 3.895868589179579e-05, + "loss": 0.199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06265155971050262, + "step": 2450, + "valid_targets_mean": 712.9, + "valid_targets_min": 465 + }, + { + "epoch": 1.3533627342888643, + "grad_norm": 0.8523741601299851, + "learning_rate": 3.894991325702483e-05, + "loss": 0.2092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0926806852221489, + "step": 2455, + "valid_targets_mean": 1248.9, + "valid_targets_min": 465 + }, + { + "epoch": 1.3561190738699007, + "grad_norm": 0.8781097559847806, + "learning_rate": 3.894110482032309e-05, + "loss": 0.2039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07843321561813354, + "step": 2460, + "valid_targets_mean": 1212.0, + "valid_targets_min": 527 + }, + { + "epoch": 1.358875413450937, + "grad_norm": 1.0781982510641095, + "learning_rate": 3.893226059833227e-05, + "loss": 0.2015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10591191053390503, + "step": 2465, + "valid_targets_mean": 1509.4, + "valid_targets_min": 561 + }, + { + "epoch": 1.3616317530319735, + "grad_norm": 0.8946480956412388, + "learning_rate": 3.892338060776172e-05, + "loss": 0.1917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08955897390842438, + "step": 2470, + "valid_targets_mean": 1309.4, + "valid_targets_min": 466 + }, + { + "epoch": 1.3643880926130099, + "grad_norm": 1.115540246875455, + "learning_rate": 3.891446486538831e-05, + "loss": 0.1936, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1010567843914032, + "step": 2475, + "valid_targets_mean": 1201.4, + "valid_targets_min": 574 + }, + { + "epoch": 1.3671444321940462, + "grad_norm": 0.8961931421083007, + "learning_rate": 3.890551338805651e-05, + "loss": 0.2097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09098277986049652, + "step": 2480, + "valid_targets_mean": 1237.0, + "valid_targets_min": 929 + }, + { + "epoch": 1.3699007717750826, + "grad_norm": 1.1595930789749056, + "learning_rate": 3.8896526192678265e-05, + "loss": 0.2022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09627296030521393, + "step": 2485, + "valid_targets_mean": 1202.0, + "valid_targets_min": 732 + }, + { + "epoch": 1.372657111356119, + "grad_norm": 0.8245236906667923, + "learning_rate": 3.8887503296233034e-05, + "loss": 0.1933, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0890670046210289, + "step": 2490, + "valid_targets_mean": 1180.6, + "valid_targets_min": 581 + }, + { + "epoch": 1.3754134509371554, + "grad_norm": 0.7255366563375965, + "learning_rate": 3.8878444715767686e-05, + "loss": 0.1908, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11314450949430466, + "step": 2495, + "valid_targets_mean": 1458.8, + "valid_targets_min": 1090 + }, + { + "epoch": 1.3781697905181918, + "grad_norm": 0.8617578620652313, + "learning_rate": 3.886935046839654e-05, + "loss": 0.1871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08164426684379578, + "step": 2500, + "valid_targets_mean": 907.9, + "valid_targets_min": 468 + }, + { + "epoch": 1.3809261300992282, + "grad_norm": 0.779833892231574, + "learning_rate": 3.886022057130129e-05, + "loss": 0.2014, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10578106343746185, + "step": 2505, + "valid_targets_mean": 1355.2, + "valid_targets_min": 492 + }, + { + "epoch": 1.3836824696802645, + "grad_norm": 0.9156774699926661, + "learning_rate": 3.885105504173099e-05, + "loss": 0.1997, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10608412325382233, + "step": 2510, + "valid_targets_mean": 1405.4, + "valid_targets_min": 704 + }, + { + "epoch": 1.386438809261301, + "grad_norm": 0.7509111792443679, + "learning_rate": 3.8841853897002e-05, + "loss": 0.2071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0969257801771164, + "step": 2515, + "valid_targets_mean": 1271.2, + "valid_targets_min": 551 + }, + { + "epoch": 1.3891951488423373, + "grad_norm": 0.858635578004762, + "learning_rate": 3.8832617154497974e-05, + "loss": 0.1997, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10384068638086319, + "step": 2520, + "valid_targets_mean": 1294.6, + "valid_targets_min": 582 + }, + { + "epoch": 1.3919514884233737, + "grad_norm": 0.8201102325910905, + "learning_rate": 3.882334483166983e-05, + "loss": 0.202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08824214339256287, + "step": 2525, + "valid_targets_mean": 1146.5, + "valid_targets_min": 637 + }, + { + "epoch": 1.39470782800441, + "grad_norm": 1.1750980173648156, + "learning_rate": 3.881403694603567e-05, + "loss": 0.2055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10282999277114868, + "step": 2530, + "valid_targets_mean": 1311.0, + "valid_targets_min": 911 + }, + { + "epoch": 1.3974641675854464, + "grad_norm": 0.5519675554776353, + "learning_rate": 3.880469351518086e-05, + "loss": 0.297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14094294607639313, + "step": 2535, + "valid_targets_mean": 4293.2, + "valid_targets_min": 464 + }, + { + "epoch": 1.4002205071664828, + "grad_norm": 0.5822758443210307, + "learning_rate": 3.8795314556757835e-05, + "loss": 0.3244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.182998389005661, + "step": 2540, + "valid_targets_mean": 4942.9, + "valid_targets_min": 551 + }, + { + "epoch": 1.4029768467475192, + "grad_norm": 0.4992292614369356, + "learning_rate": 3.878590008848621e-05, + "loss": 0.3115, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12679068744182587, + "step": 2545, + "valid_targets_mean": 4082.9, + "valid_targets_min": 1827 + }, + { + "epoch": 1.4057331863285556, + "grad_norm": 0.5564268309312894, + "learning_rate": 3.877645012815267e-05, + "loss": 0.3116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12778586149215698, + "step": 2550, + "valid_targets_mean": 3299.1, + "valid_targets_min": 1441 + }, + { + "epoch": 1.4084895259095922, + "grad_norm": 0.5690393801270861, + "learning_rate": 3.8766964693610945e-05, + "loss": 0.3065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18931175768375397, + "step": 2555, + "valid_targets_mean": 4508.8, + "valid_targets_min": 1547 + }, + { + "epoch": 1.4112458654906286, + "grad_norm": 0.4990821903757267, + "learning_rate": 3.875744380278181e-05, + "loss": 0.3189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15653735399246216, + "step": 2560, + "valid_targets_mean": 5931.4, + "valid_targets_min": 1511 + }, + { + "epoch": 1.414002205071665, + "grad_norm": 0.44878019323219615, + "learning_rate": 3.8747887473653004e-05, + "loss": 0.2981, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16134873032569885, + "step": 2565, + "valid_targets_mean": 7490.9, + "valid_targets_min": 2040 + }, + { + "epoch": 1.4167585446527013, + "grad_norm": 0.5025312848612653, + "learning_rate": 3.8738295724279226e-05, + "loss": 0.3073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09999815374612808, + "step": 2570, + "valid_targets_mean": 2932.9, + "valid_targets_min": 290 + }, + { + "epoch": 1.4195148842337377, + "grad_norm": 0.5093099024395914, + "learning_rate": 3.8728668572782103e-05, + "loss": 0.3115, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16906969249248505, + "step": 2575, + "valid_targets_mean": 5701.0, + "valid_targets_min": 1290 + }, + { + "epoch": 1.422271223814774, + "grad_norm": 0.574200275757602, + "learning_rate": 3.871900603735015e-05, + "loss": 0.3151, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1463409662246704, + "step": 2580, + "valid_targets_mean": 4178.8, + "valid_targets_min": 1054 + }, + { + "epoch": 1.4250275633958105, + "grad_norm": 0.6645711614216067, + "learning_rate": 3.870930813623871e-05, + "loss": 0.3099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16637417674064636, + "step": 2585, + "valid_targets_mean": 3651.5, + "valid_targets_min": 968 + }, + { + "epoch": 1.4277839029768469, + "grad_norm": 0.5271630004059709, + "learning_rate": 3.8699574887769953e-05, + "loss": 0.3055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15751433372497559, + "step": 2590, + "valid_targets_mean": 5212.2, + "valid_targets_min": 1273 + }, + { + "epoch": 1.4305402425578833, + "grad_norm": 0.5714782326636636, + "learning_rate": 3.8689806310332855e-05, + "loss": 0.3122, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16233137249946594, + "step": 2595, + "valid_targets_mean": 3800.2, + "valid_targets_min": 956 + }, + { + "epoch": 1.4332965821389196, + "grad_norm": 0.568859508509162, + "learning_rate": 3.868000242238309e-05, + "loss": 0.3027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17392724752426147, + "step": 2600, + "valid_targets_mean": 4258.0, + "valid_targets_min": 1441 + }, + { + "epoch": 1.436052921719956, + "grad_norm": 0.5712195247802192, + "learning_rate": 3.867016324244308e-05, + "loss": 0.3161, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12728551030158997, + "step": 2605, + "valid_targets_mean": 3077.5, + "valid_targets_min": 924 + }, + { + "epoch": 1.4388092613009924, + "grad_norm": 0.6450163916700083, + "learning_rate": 3.8660288789101925e-05, + "loss": 0.3157, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16607390344142914, + "step": 2610, + "valid_targets_mean": 3072.4, + "valid_targets_min": 1407 + }, + { + "epoch": 1.4415656008820288, + "grad_norm": 0.8384574355541901, + "learning_rate": 3.865037908101535e-05, + "loss": 0.3109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14625436067581177, + "step": 2615, + "valid_targets_mean": 2772.9, + "valid_targets_min": 382 + }, + { + "epoch": 1.4443219404630652, + "grad_norm": 0.619847351166825, + "learning_rate": 3.864043413690569e-05, + "loss": 0.3225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1614183932542801, + "step": 2620, + "valid_targets_mean": 3333.5, + "valid_targets_min": 1021 + }, + { + "epoch": 1.4470782800441016, + "grad_norm": 0.7355692683457918, + "learning_rate": 3.8630453975561856e-05, + "loss": 0.3156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11184807121753693, + "step": 2625, + "valid_targets_mean": 1783.8, + "valid_targets_min": 541 + }, + { + "epoch": 1.449834619625138, + "grad_norm": 0.6175488623984221, + "learning_rate": 3.8620438615839286e-05, + "loss": 0.3135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14034965634346008, + "step": 2630, + "valid_targets_mean": 3691.0, + "valid_targets_min": 1207 + }, + { + "epoch": 1.4525909592061743, + "grad_norm": 0.6118930225380195, + "learning_rate": 3.8610388076659944e-05, + "loss": 0.3163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14856155216693878, + "step": 2635, + "valid_targets_mean": 3065.6, + "valid_targets_min": 767 + }, + { + "epoch": 1.4553472987872107, + "grad_norm": 1.3118844887551553, + "learning_rate": 3.8600302377012226e-05, + "loss": 0.3034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19144447147846222, + "step": 2640, + "valid_targets_mean": 3770.9, + "valid_targets_min": 1544 + }, + { + "epoch": 1.458103638368247, + "grad_norm": 0.5416233736681149, + "learning_rate": 3.8590181535950966e-05, + "loss": 0.2816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11874759942293167, + "step": 2645, + "valid_targets_mean": 3893.1, + "valid_targets_min": 1633 + }, + { + "epoch": 1.4608599779492835, + "grad_norm": 0.6072767677993843, + "learning_rate": 3.858002557259741e-05, + "loss": 0.3106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14717033505439758, + "step": 2650, + "valid_targets_mean": 3006.1, + "valid_targets_min": 1381 + }, + { + "epoch": 1.4636163175303198, + "grad_norm": 0.6330162570264669, + "learning_rate": 3.856983450613912e-05, + "loss": 0.3106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15561364591121674, + "step": 2655, + "valid_targets_mean": 3179.0, + "valid_targets_min": 418 + }, + { + "epoch": 1.4663726571113562, + "grad_norm": 0.5373576153270534, + "learning_rate": 3.855960835583003e-05, + "loss": 0.3173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15485015511512756, + "step": 2660, + "valid_targets_mean": 3881.4, + "valid_targets_min": 1596 + }, + { + "epoch": 1.4691289966923926, + "grad_norm": 0.6609514918851265, + "learning_rate": 3.854934714099031e-05, + "loss": 0.295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1196301132440567, + "step": 2665, + "valid_targets_mean": 2007.9, + "valid_targets_min": 994 + }, + { + "epoch": 1.471885336273429, + "grad_norm": 0.5626860818912256, + "learning_rate": 3.8539050881006404e-05, + "loss": 0.2937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14879678189754486, + "step": 2670, + "valid_targets_mean": 3667.6, + "valid_targets_min": 1738 + }, + { + "epoch": 1.4746416758544654, + "grad_norm": 0.7486122882795879, + "learning_rate": 3.852871959533096e-05, + "loss": 0.3131, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15494510531425476, + "step": 2675, + "valid_targets_mean": 3595.1, + "valid_targets_min": 612 + }, + { + "epoch": 1.4773980154355018, + "grad_norm": 0.5768187521277942, + "learning_rate": 3.8518353303482796e-05, + "loss": 0.3032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13747937977313995, + "step": 2680, + "valid_targets_mean": 2771.0, + "valid_targets_min": 1190 + }, + { + "epoch": 1.4801543550165381, + "grad_norm": 0.5003637638389149, + "learning_rate": 3.850795202504688e-05, + "loss": 0.289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12953521311283112, + "step": 2685, + "valid_targets_mean": 4476.0, + "valid_targets_min": 1648 + }, + { + "epoch": 1.4829106945975745, + "grad_norm": 0.6679435137533197, + "learning_rate": 3.8497515779674254e-05, + "loss": 0.3064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14474862813949585, + "step": 2690, + "valid_targets_mean": 3159.1, + "valid_targets_min": 951 + }, + { + "epoch": 1.485667034178611, + "grad_norm": 0.5497509096161187, + "learning_rate": 3.848704458708205e-05, + "loss": 0.3046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15493667125701904, + "step": 2695, + "valid_targets_mean": 3956.6, + "valid_targets_min": 1863 + }, + { + "epoch": 1.4884233737596473, + "grad_norm": 0.6627280641831833, + "learning_rate": 3.847653846705342e-05, + "loss": 0.2956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17405965924263, + "step": 2700, + "valid_targets_mean": 3484.0, + "valid_targets_min": 2153 + }, + { + "epoch": 1.4911797133406837, + "grad_norm": 0.7534730391686231, + "learning_rate": 3.846599743943749e-05, + "loss": 0.3075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12704578042030334, + "step": 2705, + "valid_targets_mean": 2673.0, + "valid_targets_min": 1094 + }, + { + "epoch": 1.49393605292172, + "grad_norm": 0.5959511945028271, + "learning_rate": 3.845542152414934e-05, + "loss": 0.3074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1465003937482834, + "step": 2710, + "valid_targets_mean": 3583.8, + "valid_targets_min": 885 + }, + { + "epoch": 1.4966923925027564, + "grad_norm": 0.5000831023666354, + "learning_rate": 3.844481074116998e-05, + "loss": 0.3022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15734559297561646, + "step": 2715, + "valid_targets_mean": 5273.2, + "valid_targets_min": 1686 + }, + { + "epoch": 1.4994487320837928, + "grad_norm": 0.5688626511323461, + "learning_rate": 3.8434165110546287e-05, + "loss": 0.2914, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17321285605430603, + "step": 2720, + "valid_targets_mean": 3796.8, + "valid_targets_min": 1682 + }, + { + "epoch": 1.5022050716648292, + "grad_norm": 0.598400714391043, + "learning_rate": 3.842348465239096e-05, + "loss": 0.3022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15532225370407104, + "step": 2725, + "valid_targets_mean": 3432.5, + "valid_targets_min": 1640 + }, + { + "epoch": 1.5049614112458656, + "grad_norm": 0.7067302633805036, + "learning_rate": 3.841276938688251e-05, + "loss": 0.302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20367062091827393, + "step": 2730, + "valid_targets_mean": 4120.2, + "valid_targets_min": 700 + }, + { + "epoch": 1.507717750826902, + "grad_norm": 0.6166030378685813, + "learning_rate": 3.8402019334265216e-05, + "loss": 0.294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1630454659461975, + "step": 2735, + "valid_targets_mean": 3153.4, + "valid_targets_min": 577 + }, + { + "epoch": 1.5104740904079383, + "grad_norm": 0.6055708088351399, + "learning_rate": 3.839123451484907e-05, + "loss": 0.306, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.184973806142807, + "step": 2740, + "valid_targets_mean": 4260.0, + "valid_targets_min": 2219 + }, + { + "epoch": 1.5132304299889747, + "grad_norm": 0.5610527696823002, + "learning_rate": 3.838041494900975e-05, + "loss": 0.3002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1724320650100708, + "step": 2745, + "valid_targets_mean": 4229.1, + "valid_targets_min": 1825 + }, + { + "epoch": 1.515986769570011, + "grad_norm": 0.5483101275519798, + "learning_rate": 3.836956065718859e-05, + "loss": 0.2898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14666646718978882, + "step": 2750, + "valid_targets_mean": 3777.9, + "valid_targets_min": 2010 + }, + { + "epoch": 1.5187431091510475, + "grad_norm": 0.5520755590663621, + "learning_rate": 3.8358671659892507e-05, + "loss": 0.2849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09732802212238312, + "step": 2755, + "valid_targets_mean": 2502.8, + "valid_targets_min": 734 + }, + { + "epoch": 1.5214994487320839, + "grad_norm": 0.8804752415139114, + "learning_rate": 3.8347747977694014e-05, + "loss": 0.3127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1818457841873169, + "step": 2760, + "valid_targets_mean": 3351.5, + "valid_targets_min": 1697 + }, + { + "epoch": 1.5242557883131203, + "grad_norm": 0.6522328924191998, + "learning_rate": 3.8336789631231136e-05, + "loss": 0.2934, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.123630590736866, + "step": 2765, + "valid_targets_mean": 2532.9, + "valid_targets_min": 1377 + }, + { + "epoch": 1.5270121278941566, + "grad_norm": 0.6779409418508127, + "learning_rate": 3.832579664120741e-05, + "loss": 0.3024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18511353433132172, + "step": 2770, + "valid_targets_mean": 3187.2, + "valid_targets_min": 2106 + }, + { + "epoch": 1.529768467475193, + "grad_norm": 0.6012563869710313, + "learning_rate": 3.83147690283918e-05, + "loss": 0.3063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1612846851348877, + "step": 2775, + "valid_targets_mean": 3775.5, + "valid_targets_min": 1896 + }, + { + "epoch": 1.5325248070562294, + "grad_norm": 0.5690634887605872, + "learning_rate": 3.830370681361869e-05, + "loss": 0.2975, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19814658164978027, + "step": 2780, + "valid_targets_mean": 4860.4, + "valid_targets_min": 1760 + }, + { + "epoch": 1.5352811466372658, + "grad_norm": 0.6329742353586345, + "learning_rate": 3.829261001778785e-05, + "loss": 0.2969, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1973462998867035, + "step": 2785, + "valid_targets_mean": 3965.0, + "valid_targets_min": 1967 + }, + { + "epoch": 1.5380374862183022, + "grad_norm": 0.6087340341913123, + "learning_rate": 3.8281478661864365e-05, + "loss": 0.3089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21505898237228394, + "step": 2790, + "valid_targets_mean": 4136.6, + "valid_targets_min": 1647 + }, + { + "epoch": 1.5407938257993385, + "grad_norm": 0.5282892160131334, + "learning_rate": 3.8270312766878624e-05, + "loss": 0.2845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12343691289424896, + "step": 2795, + "valid_targets_mean": 3405.9, + "valid_targets_min": 1902 + }, + { + "epoch": 1.543550165380375, + "grad_norm": 0.5796762738895086, + "learning_rate": 3.825911235392628e-05, + "loss": 0.2849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16388481855392456, + "step": 2800, + "valid_targets_mean": 3465.0, + "valid_targets_min": 2261 + }, + { + "epoch": 1.5463065049614113, + "grad_norm": 0.5139534724696003, + "learning_rate": 3.824787744416818e-05, + "loss": 0.2826, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13608059287071228, + "step": 2805, + "valid_targets_mean": 4239.8, + "valid_targets_min": 902 + }, + { + "epoch": 1.5490628445424477, + "grad_norm": 0.6362714675495325, + "learning_rate": 3.823660805883036e-05, + "loss": 0.2961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1707872748374939, + "step": 2810, + "valid_targets_mean": 3418.5, + "valid_targets_min": 742 + }, + { + "epoch": 1.551819184123484, + "grad_norm": 0.5723746467418398, + "learning_rate": 3.822530421920399e-05, + "loss": 0.2976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14277589321136475, + "step": 2815, + "valid_targets_mean": 3846.0, + "valid_targets_min": 1355 + }, + { + "epoch": 1.5545755237045205, + "grad_norm": 0.6141266973713077, + "learning_rate": 3.8213965946645324e-05, + "loss": 0.2963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14834712445735931, + "step": 2820, + "valid_targets_mean": 3270.1, + "valid_targets_min": 1242 + }, + { + "epoch": 1.5573318632855568, + "grad_norm": 0.5480986819499922, + "learning_rate": 3.8202593262575674e-05, + "loss": 0.294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11913686990737915, + "step": 2825, + "valid_targets_mean": 3767.4, + "valid_targets_min": 2073 + }, + { + "epoch": 1.5600882028665932, + "grad_norm": 0.6030313982981457, + "learning_rate": 3.819118618848137e-05, + "loss": 0.3003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1374896615743637, + "step": 2830, + "valid_targets_mean": 3248.2, + "valid_targets_min": 906 + }, + { + "epoch": 1.5628445424476296, + "grad_norm": 0.5469132822040982, + "learning_rate": 3.8179744745913724e-05, + "loss": 0.3015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12800200283527374, + "step": 2835, + "valid_targets_mean": 3214.2, + "valid_targets_min": 616 + }, + { + "epoch": 1.565600882028666, + "grad_norm": 0.6844844918360014, + "learning_rate": 3.816826895648895e-05, + "loss": 0.2973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12773922085762024, + "step": 2840, + "valid_targets_mean": 2117.8, + "valid_targets_min": 819 + }, + { + "epoch": 1.5683572216097024, + "grad_norm": 0.6293446335629433, + "learning_rate": 3.815675884188818e-05, + "loss": 0.2855, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15198874473571777, + "step": 2845, + "valid_targets_mean": 3294.6, + "valid_targets_min": 1143 + }, + { + "epoch": 1.5711135611907387, + "grad_norm": 0.5272111241298029, + "learning_rate": 3.81452144238574e-05, + "loss": 0.2796, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1474853754043579, + "step": 2850, + "valid_targets_mean": 4368.0, + "valid_targets_min": 1191 + }, + { + "epoch": 1.5738699007717751, + "grad_norm": 0.5828307124768471, + "learning_rate": 3.813363572420738e-05, + "loss": 0.2873, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15340495109558105, + "step": 2855, + "valid_targets_mean": 3497.1, + "valid_targets_min": 289 + }, + { + "epoch": 1.5766262403528115, + "grad_norm": 0.5350548957161192, + "learning_rate": 3.8122022764813674e-05, + "loss": 0.2779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11237572133541107, + "step": 2860, + "valid_targets_mean": 3379.6, + "valid_targets_min": 989 + }, + { + "epoch": 1.579382579933848, + "grad_norm": 0.6914793533534679, + "learning_rate": 3.811037556761656e-05, + "loss": 0.2779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12102369964122772, + "step": 2865, + "valid_targets_mean": 1870.4, + "valid_targets_min": 344 + }, + { + "epoch": 1.5821389195148843, + "grad_norm": 0.6553918872943303, + "learning_rate": 3.8098694154621004e-05, + "loss": 0.2801, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14969420433044434, + "step": 2870, + "valid_targets_mean": 3110.2, + "valid_targets_min": 2053 + }, + { + "epoch": 1.5848952590959207, + "grad_norm": 0.6273319707062087, + "learning_rate": 3.808697854789661e-05, + "loss": 0.2931, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20834359526634216, + "step": 2875, + "valid_targets_mean": 4253.2, + "valid_targets_min": 2066 + }, + { + "epoch": 1.587651598676957, + "grad_norm": 0.5931533230675375, + "learning_rate": 3.8075228769577586e-05, + "loss": 0.3086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12589780986309052, + "step": 2880, + "valid_targets_mean": 3006.6, + "valid_targets_min": 1502 + }, + { + "epoch": 1.5904079382579934, + "grad_norm": 0.7415614110660992, + "learning_rate": 3.806344484186271e-05, + "loss": 0.2871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1705782115459442, + "step": 2885, + "valid_targets_mean": 2874.1, + "valid_targets_min": 800 + }, + { + "epoch": 1.5931642778390298, + "grad_norm": 0.5732708229409031, + "learning_rate": 3.805162678701526e-05, + "loss": 0.2917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13199031352996826, + "step": 2890, + "valid_targets_mean": 3433.6, + "valid_targets_min": 1415 + }, + { + "epoch": 1.5959206174200662, + "grad_norm": 0.5176563670786888, + "learning_rate": 3.8039774627363e-05, + "loss": 0.2814, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15882933139801025, + "step": 2895, + "valid_targets_mean": 4436.1, + "valid_targets_min": 1292 + }, + { + "epoch": 1.5986769570011026, + "grad_norm": 0.47071644903079024, + "learning_rate": 3.802788838529813e-05, + "loss": 0.2811, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12790828943252563, + "step": 2900, + "valid_targets_mean": 5184.6, + "valid_targets_min": 1521 + }, + { + "epoch": 1.601433296582139, + "grad_norm": 0.5432529574110507, + "learning_rate": 3.801596808327724e-05, + "loss": 0.2876, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13238166272640228, + "step": 2905, + "valid_targets_mean": 4194.6, + "valid_targets_min": 1789 + }, + { + "epoch": 1.6041896361631753, + "grad_norm": 0.6377763508964951, + "learning_rate": 3.800401374382127e-05, + "loss": 0.2769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11496248096227646, + "step": 2910, + "valid_targets_mean": 2431.6, + "valid_targets_min": 1470 + }, + { + "epoch": 1.6069459757442117, + "grad_norm": 0.534028792597784, + "learning_rate": 3.799202538951546e-05, + "loss": 0.2908, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10071459412574768, + "step": 2915, + "valid_targets_mean": 2841.1, + "valid_targets_min": 1282 + }, + { + "epoch": 1.609702315325248, + "grad_norm": 0.6502975949413557, + "learning_rate": 3.798000304300932e-05, + "loss": 0.2839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15787151455879211, + "step": 2920, + "valid_targets_mean": 3547.6, + "valid_targets_min": 1422 + }, + { + "epoch": 1.6124586549062845, + "grad_norm": 0.5806330125095023, + "learning_rate": 3.796794672701658e-05, + "loss": 0.2953, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13371174037456512, + "step": 2925, + "valid_targets_mean": 2877.2, + "valid_targets_min": 1210 + }, + { + "epoch": 1.6152149944873209, + "grad_norm": 0.6143067160650585, + "learning_rate": 3.795585646431516e-05, + "loss": 0.3019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14754506945610046, + "step": 2930, + "valid_targets_mean": 3072.8, + "valid_targets_min": 1139 + }, + { + "epoch": 1.6179713340683572, + "grad_norm": 0.7729844312574835, + "learning_rate": 3.794373227774708e-05, + "loss": 0.2938, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1830815225839615, + "step": 2935, + "valid_targets_mean": 2247.0, + "valid_targets_min": 619 + }, + { + "epoch": 1.6207276736493936, + "grad_norm": 0.5909043872918606, + "learning_rate": 3.7931574190218504e-05, + "loss": 0.3136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10417908430099487, + "step": 2940, + "valid_targets_mean": 2174.6, + "valid_targets_min": 447 + }, + { + "epoch": 1.62348401323043, + "grad_norm": 0.531183883219299, + "learning_rate": 3.791938222469959e-05, + "loss": 0.2844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14901182055473328, + "step": 2945, + "valid_targets_mean": 4156.4, + "valid_targets_min": 2452 + }, + { + "epoch": 1.6262403528114664, + "grad_norm": 0.5668769159184646, + "learning_rate": 3.790715640422455e-05, + "loss": 0.3036, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12196837365627289, + "step": 2950, + "valid_targets_mean": 3091.2, + "valid_targets_min": 837 + }, + { + "epoch": 1.6289966923925028, + "grad_norm": 0.6471853973017421, + "learning_rate": 3.789489675189152e-05, + "loss": 0.3073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19502410292625427, + "step": 2955, + "valid_targets_mean": 3410.5, + "valid_targets_min": 794 + }, + { + "epoch": 1.6317530319735392, + "grad_norm": 0.6044396518608646, + "learning_rate": 3.788260329086258e-05, + "loss": 0.2862, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14365798234939575, + "step": 2960, + "valid_targets_mean": 3586.6, + "valid_targets_min": 808 + }, + { + "epoch": 1.6345093715545755, + "grad_norm": 0.5983478451540356, + "learning_rate": 3.787027604436367e-05, + "loss": 0.2918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16552947461605072, + "step": 2965, + "valid_targets_mean": 3692.8, + "valid_targets_min": 1948 + }, + { + "epoch": 1.637265711135612, + "grad_norm": 0.5793015368211132, + "learning_rate": 3.785791503568457e-05, + "loss": 0.2943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12275514006614685, + "step": 2970, + "valid_targets_mean": 3656.9, + "valid_targets_min": 1909 + }, + { + "epoch": 1.6400220507166483, + "grad_norm": 0.6205430484941757, + "learning_rate": 3.784552028817884e-05, + "loss": 0.2978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13549336791038513, + "step": 2975, + "valid_targets_mean": 2981.6, + "valid_targets_min": 1511 + }, + { + "epoch": 1.6427783902976847, + "grad_norm": 0.6685303305208627, + "learning_rate": 3.7833091825263785e-05, + "loss": 0.2995, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15516670048236847, + "step": 2980, + "valid_targets_mean": 2484.5, + "valid_targets_min": 450 + }, + { + "epoch": 1.645534729878721, + "grad_norm": 0.5171997838601483, + "learning_rate": 3.7820629670420406e-05, + "loss": 0.2731, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1288493126630783, + "step": 2985, + "valid_targets_mean": 3432.0, + "valid_targets_min": 2554 + }, + { + "epoch": 1.6482910694597575, + "grad_norm": 0.5778911176294369, + "learning_rate": 3.780813384719336e-05, + "loss": 0.2794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.143443763256073, + "step": 2990, + "valid_targets_mean": 3475.6, + "valid_targets_min": 1202 + }, + { + "epoch": 1.6510474090407938, + "grad_norm": 0.543251181971293, + "learning_rate": 3.779560437919091e-05, + "loss": 0.2835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13294801115989685, + "step": 2995, + "valid_targets_mean": 3600.8, + "valid_targets_min": 2159 + }, + { + "epoch": 1.6538037486218302, + "grad_norm": 0.7062004301336318, + "learning_rate": 3.7783041290084904e-05, + "loss": 0.2827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14826326072216034, + "step": 3000, + "valid_targets_mean": 2471.4, + "valid_targets_min": 808 + }, + { + "epoch": 1.6565600882028666, + "grad_norm": 0.5307176582539255, + "learning_rate": 3.777044460361067e-05, + "loss": 0.2812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11649367213249207, + "step": 3005, + "valid_targets_mean": 3015.4, + "valid_targets_min": 990 + }, + { + "epoch": 1.659316427783903, + "grad_norm": 0.5903841404704616, + "learning_rate": 3.775781434356705e-05, + "loss": 0.2936, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13989458978176117, + "step": 3010, + "valid_targets_mean": 3459.4, + "valid_targets_min": 318 + }, + { + "epoch": 1.6620727673649394, + "grad_norm": 0.5454820718282699, + "learning_rate": 3.77451505338163e-05, + "loss": 0.2805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15865656733512878, + "step": 3015, + "valid_targets_mean": 4340.6, + "valid_targets_min": 2582 + }, + { + "epoch": 1.6648291069459757, + "grad_norm": 0.7365927968655802, + "learning_rate": 3.773245319828405e-05, + "loss": 0.2923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15623635053634644, + "step": 3020, + "valid_targets_mean": 2988.0, + "valid_targets_min": 1117 + }, + { + "epoch": 1.6675854465270121, + "grad_norm": 0.5034768898885823, + "learning_rate": 3.771972236095929e-05, + "loss": 0.2854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12304627895355225, + "step": 3025, + "valid_targets_mean": 3749.8, + "valid_targets_min": 1890 + }, + { + "epoch": 1.6703417861080485, + "grad_norm": 0.5912967354789794, + "learning_rate": 3.77069580458943e-05, + "loss": 0.2751, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11723654717206955, + "step": 3030, + "valid_targets_mean": 2966.2, + "valid_targets_min": 1793 + }, + { + "epoch": 1.673098125689085, + "grad_norm": 0.5658697318586998, + "learning_rate": 3.76941602772046e-05, + "loss": 0.2671, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11708976328372955, + "step": 3035, + "valid_targets_mean": 3508.5, + "valid_targets_min": 773 + }, + { + "epoch": 1.6758544652701213, + "grad_norm": 0.8201967512263265, + "learning_rate": 3.768132907906893e-05, + "loss": 0.2749, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15111519396305084, + "step": 3040, + "valid_targets_mean": 3232.2, + "valid_targets_min": 1599 + }, + { + "epoch": 1.6786108048511577, + "grad_norm": 0.5598462364951848, + "learning_rate": 3.766846447572916e-05, + "loss": 0.2769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13623324036598206, + "step": 3045, + "valid_targets_mean": 3505.6, + "valid_targets_min": 776 + }, + { + "epoch": 1.681367144432194, + "grad_norm": 0.5393900554659449, + "learning_rate": 3.765556649149031e-05, + "loss": 0.2738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17005959153175354, + "step": 3050, + "valid_targets_mean": 4465.2, + "valid_targets_min": 1866 + }, + { + "epoch": 1.6841234840132304, + "grad_norm": 0.5855693596934327, + "learning_rate": 3.7642635150720426e-05, + "loss": 0.268, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14446868002414703, + "step": 3055, + "valid_targets_mean": 3549.2, + "valid_targets_min": 735 + }, + { + "epoch": 1.6868798235942668, + "grad_norm": 0.572718727010268, + "learning_rate": 3.7629670477850616e-05, + "loss": 0.2744, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12344960868358612, + "step": 3060, + "valid_targets_mean": 2900.2, + "valid_targets_min": 1470 + }, + { + "epoch": 1.6896361631753032, + "grad_norm": 0.652231078861618, + "learning_rate": 3.761667249737491e-05, + "loss": 0.2775, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10677766054868698, + "step": 3065, + "valid_targets_mean": 2116.2, + "valid_targets_min": 373 + }, + { + "epoch": 1.6923925027563396, + "grad_norm": 0.6153586225485684, + "learning_rate": 3.760364123385033e-05, + "loss": 0.2993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12259036302566528, + "step": 3070, + "valid_targets_mean": 2585.5, + "valid_targets_min": 952 + }, + { + "epoch": 1.695148842337376, + "grad_norm": 0.6672297776626556, + "learning_rate": 3.759057671189673e-05, + "loss": 0.2811, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1338804066181183, + "step": 3075, + "valid_targets_mean": 2983.9, + "valid_targets_min": 1505 + }, + { + "epoch": 1.6979051819184123, + "grad_norm": 0.5774052439448606, + "learning_rate": 3.757747895619681e-05, + "loss": 0.2815, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1256086528301239, + "step": 3080, + "valid_targets_mean": 3333.4, + "valid_targets_min": 633 + }, + { + "epoch": 1.7006615214994487, + "grad_norm": 0.6063563876622791, + "learning_rate": 3.756434799149607e-05, + "loss": 0.2713, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15054842829704285, + "step": 3085, + "valid_targets_mean": 3836.8, + "valid_targets_min": 1483 + }, + { + "epoch": 1.703417861080485, + "grad_norm": 0.5641962787421195, + "learning_rate": 3.7551183842602735e-05, + "loss": 0.2783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13165020942687988, + "step": 3090, + "valid_targets_mean": 3521.6, + "valid_targets_min": 1797 + }, + { + "epoch": 1.7061742006615215, + "grad_norm": 0.5688049179984976, + "learning_rate": 3.7537986534387754e-05, + "loss": 0.2862, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13853976130485535, + "step": 3095, + "valid_targets_mean": 3853.1, + "valid_targets_min": 2662 + }, + { + "epoch": 1.7089305402425579, + "grad_norm": 0.55518422550932, + "learning_rate": 3.752475609178468e-05, + "loss": 0.2874, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14337566494941711, + "step": 3100, + "valid_targets_mean": 4088.1, + "valid_targets_min": 2230 + }, + { + "epoch": 1.7116868798235942, + "grad_norm": 0.5605962914104132, + "learning_rate": 3.75114925397897e-05, + "loss": 0.279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11106985807418823, + "step": 3105, + "valid_targets_mean": 3464.5, + "valid_targets_min": 1628 + }, + { + "epoch": 1.7144432194046306, + "grad_norm": 0.5835826698113092, + "learning_rate": 3.7498195903461556e-05, + "loss": 0.2797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16153426468372345, + "step": 3110, + "valid_targets_mean": 4436.0, + "valid_targets_min": 830 + }, + { + "epoch": 1.717199558985667, + "grad_norm": 0.530558928361706, + "learning_rate": 3.748486620792147e-05, + "loss": 0.2781, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11890096962451935, + "step": 3115, + "valid_targets_mean": 3477.5, + "valid_targets_min": 1715 + }, + { + "epoch": 1.7199558985667034, + "grad_norm": 0.5334921991568702, + "learning_rate": 3.747150347835314e-05, + "loss": 0.2231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09813651442527771, + "step": 3120, + "valid_targets_mean": 2246.1, + "valid_targets_min": 331 + }, + { + "epoch": 1.7227122381477398, + "grad_norm": 0.38463255236215443, + "learning_rate": 3.7458107740002676e-05, + "loss": 0.133, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.062398798763751984, + "step": 3125, + "valid_targets_mean": 3604.4, + "valid_targets_min": 708 + }, + { + "epoch": 1.7254685777287762, + "grad_norm": 0.4594847604786553, + "learning_rate": 3.744467901817854e-05, + "loss": 0.1167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06351009756326675, + "step": 3130, + "valid_targets_mean": 2933.4, + "valid_targets_min": 729 + }, + { + "epoch": 1.7282249173098125, + "grad_norm": 0.43774323809052196, + "learning_rate": 3.743121733825152e-05, + "loss": 0.1204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10130656510591507, + "step": 3135, + "valid_targets_mean": 2718.9, + "valid_targets_min": 921 + }, + { + "epoch": 1.730981256890849, + "grad_norm": 0.3879326381580189, + "learning_rate": 3.741772272565468e-05, + "loss": 0.105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04767099767923355, + "step": 3140, + "valid_targets_mean": 2827.9, + "valid_targets_min": 813 + }, + { + "epoch": 1.7337375964718853, + "grad_norm": 0.44787717866328525, + "learning_rate": 3.7404195205883274e-05, + "loss": 0.1193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06180643290281296, + "step": 3145, + "valid_targets_mean": 2883.1, + "valid_targets_min": 854 + }, + { + "epoch": 1.7364939360529217, + "grad_norm": 0.8092623953343431, + "learning_rate": 3.7390634804494757e-05, + "loss": 0.1682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06640288233757019, + "step": 3150, + "valid_targets_mean": 1034.4, + "valid_targets_min": 536 + }, + { + "epoch": 1.739250275633958, + "grad_norm": 0.39634308793757034, + "learning_rate": 3.737704154710868e-05, + "loss": 0.1304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05220836028456688, + "step": 3155, + "valid_targets_mean": 2992.8, + "valid_targets_min": 915 + }, + { + "epoch": 1.7420066152149944, + "grad_norm": 0.398855581861587, + "learning_rate": 3.73634154594067e-05, + "loss": 0.1259, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06232686713337898, + "step": 3160, + "valid_targets_mean": 3356.0, + "valid_targets_min": 2281 + }, + { + "epoch": 1.7447629547960308, + "grad_norm": 0.34849410485224014, + "learning_rate": 3.734975656713248e-05, + "loss": 0.1581, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04277760535478592, + "step": 3165, + "valid_targets_mean": 3776.1, + "valid_targets_min": 958 + }, + { + "epoch": 1.7475192943770672, + "grad_norm": 0.35497742895455425, + "learning_rate": 3.733606489609166e-05, + "loss": 0.0969, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06533674150705338, + "step": 3170, + "valid_targets_mean": 4223.1, + "valid_targets_min": 2082 + }, + { + "epoch": 1.7502756339581036, + "grad_norm": 0.3783696074824077, + "learning_rate": 3.732234047215181e-05, + "loss": 0.1108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05947469174861908, + "step": 3175, + "valid_targets_mean": 4378.9, + "valid_targets_min": 3337 + }, + { + "epoch": 1.75303197353914, + "grad_norm": 0.32490633048503176, + "learning_rate": 3.730858332124239e-05, + "loss": 0.1102, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0541544184088707, + "step": 3180, + "valid_targets_mean": 3910.5, + "valid_targets_min": 515 + }, + { + "epoch": 1.7557883131201764, + "grad_norm": 0.5785208989345174, + "learning_rate": 3.729479346935468e-05, + "loss": 0.1635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06321974098682404, + "step": 3185, + "valid_targets_mean": 1878.6, + "valid_targets_min": 546 + }, + { + "epoch": 1.7585446527012127, + "grad_norm": 0.37399802390403863, + "learning_rate": 3.728097094254174e-05, + "loss": 0.1039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04874004051089287, + "step": 3190, + "valid_targets_mean": 3397.5, + "valid_targets_min": 2715 + }, + { + "epoch": 1.7613009922822491, + "grad_norm": 0.4612629046882934, + "learning_rate": 3.726711576691838e-05, + "loss": 0.1191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046907827258110046, + "step": 3195, + "valid_targets_mean": 1612.8, + "valid_targets_min": 515 + }, + { + "epoch": 1.7640573318632855, + "grad_norm": 0.3637232164367072, + "learning_rate": 3.725322796866106e-05, + "loss": 0.0992, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031437601894140244, + "step": 3200, + "valid_targets_mean": 3322.8, + "valid_targets_min": 840 + }, + { + "epoch": 1.7668136714443219, + "grad_norm": 0.4319998299066277, + "learning_rate": 3.7239307574007916e-05, + "loss": 0.1158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05722799152135849, + "step": 3205, + "valid_targets_mean": 2764.0, + "valid_targets_min": 574 + }, + { + "epoch": 1.7695700110253583, + "grad_norm": 0.4797500009376479, + "learning_rate": 3.722535460925864e-05, + "loss": 0.1218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06558194756507874, + "step": 3210, + "valid_targets_mean": 2077.9, + "valid_targets_min": 847 + }, + { + "epoch": 1.7723263506063947, + "grad_norm": 0.4982822137405986, + "learning_rate": 3.721136910077446e-05, + "loss": 0.1291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07162708789110184, + "step": 3215, + "valid_targets_mean": 2857.1, + "valid_targets_min": 770 + }, + { + "epoch": 1.775082690187431, + "grad_norm": 0.47272363088366176, + "learning_rate": 3.71973510749781e-05, + "loss": 0.1266, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051692232489585876, + "step": 3220, + "valid_targets_mean": 1736.5, + "valid_targets_min": 611 + }, + { + "epoch": 1.7778390297684674, + "grad_norm": 0.434593271551419, + "learning_rate": 3.7183300558353704e-05, + "loss": 0.1645, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09611590206623077, + "step": 3225, + "valid_targets_mean": 3113.2, + "valid_targets_min": 1665 + }, + { + "epoch": 1.7805953693495038, + "grad_norm": 0.3170482372763072, + "learning_rate": 3.716921757744682e-05, + "loss": 0.108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04649633541703224, + "step": 3230, + "valid_targets_mean": 4061.4, + "valid_targets_min": 3722 + }, + { + "epoch": 1.7833517089305402, + "grad_norm": 0.5540650133170724, + "learning_rate": 3.715510215886431e-05, + "loss": 0.118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05161561071872711, + "step": 3235, + "valid_targets_mean": 1199.0, + "valid_targets_min": 711 + }, + { + "epoch": 1.7861080485115766, + "grad_norm": 0.37866144641129434, + "learning_rate": 3.714095432927433e-05, + "loss": 0.1561, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0516941137611866, + "step": 3240, + "valid_targets_mean": 3407.8, + "valid_targets_min": 1803 + }, + { + "epoch": 1.788864388092613, + "grad_norm": 0.36236580582876166, + "learning_rate": 3.712677411540627e-05, + "loss": 0.1365, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05718575417995453, + "step": 3245, + "valid_targets_mean": 3789.4, + "valid_targets_min": 3167 + }, + { + "epoch": 1.7916207276736493, + "grad_norm": 0.7306941313330693, + "learning_rate": 3.711256154405071e-05, + "loss": 0.1482, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12909424304962158, + "step": 3250, + "valid_targets_mean": 1659.2, + "valid_targets_min": 864 + }, + { + "epoch": 1.7943770672546857, + "grad_norm": 0.41305994868946405, + "learning_rate": 3.709831664205935e-05, + "loss": 0.2978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06506998836994171, + "step": 3255, + "valid_targets_mean": 3663.0, + "valid_targets_min": 2876 + }, + { + "epoch": 1.797133406835722, + "grad_norm": 0.42288094310105934, + "learning_rate": 3.708403943634499e-05, + "loss": 0.1304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0557003915309906, + "step": 3260, + "valid_targets_mean": 3140.6, + "valid_targets_min": 828 + }, + { + "epoch": 1.7998897464167585, + "grad_norm": 0.4943632863965581, + "learning_rate": 3.706972995388143e-05, + "loss": 0.1279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10280945152044296, + "step": 3265, + "valid_targets_mean": 3309.0, + "valid_targets_min": 1794 + }, + { + "epoch": 1.8026460859977949, + "grad_norm": 0.3928981973093974, + "learning_rate": 3.705538822170348e-05, + "loss": 0.2233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07008827477693558, + "step": 3270, + "valid_targets_mean": 3527.5, + "valid_targets_min": 2771 + }, + { + "epoch": 1.8054024255788312, + "grad_norm": 0.2844313867164904, + "learning_rate": 3.704101426690686e-05, + "loss": 0.0934, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04273977503180504, + "step": 3275, + "valid_targets_mean": 4565.0, + "valid_targets_min": 2655 + }, + { + "epoch": 1.8081587651598676, + "grad_norm": 0.3555609558269117, + "learning_rate": 3.702660811664819e-05, + "loss": 0.1266, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029215069487690926, + "step": 3280, + "valid_targets_mean": 2743.9, + "valid_targets_min": 557 + }, + { + "epoch": 1.810915104740904, + "grad_norm": 0.4328327110237278, + "learning_rate": 3.70121697981449e-05, + "loss": 0.124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06079603731632233, + "step": 3285, + "valid_targets_mean": 4249.6, + "valid_targets_min": 1606 + }, + { + "epoch": 1.8136714443219404, + "grad_norm": 0.3682964002054086, + "learning_rate": 3.6997699338675184e-05, + "loss": 0.1079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0480961911380291, + "step": 3290, + "valid_targets_mean": 3533.0, + "valid_targets_min": 2163 + }, + { + "epoch": 1.8164277839029768, + "grad_norm": 0.4056843850568332, + "learning_rate": 3.698319676557799e-05, + "loss": 0.1057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0556955523788929, + "step": 3295, + "valid_targets_mean": 3057.5, + "valid_targets_min": 533 + }, + { + "epoch": 1.8191841234840131, + "grad_norm": 0.39778211928032225, + "learning_rate": 3.696866210625291e-05, + "loss": 0.1082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056053344160318375, + "step": 3300, + "valid_targets_mean": 3622.0, + "valid_targets_min": 2249 + }, + { + "epoch": 1.8219404630650495, + "grad_norm": 0.38422137766613534, + "learning_rate": 3.695409538816018e-05, + "loss": 0.1028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04674785956740379, + "step": 3305, + "valid_targets_mean": 2801.5, + "valid_targets_min": 817 + }, + { + "epoch": 1.824696802646086, + "grad_norm": 0.5826723810993907, + "learning_rate": 3.693949663882058e-05, + "loss": 0.1576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1456746608018875, + "step": 3310, + "valid_targets_mean": 1954.6, + "valid_targets_min": 859 + }, + { + "epoch": 1.8274531422271223, + "grad_norm": 0.4005149965170071, + "learning_rate": 3.692486588581543e-05, + "loss": 0.1018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0575944148004055, + "step": 3315, + "valid_targets_mean": 3564.0, + "valid_targets_min": 2709 + }, + { + "epoch": 1.8302094818081587, + "grad_norm": 0.6406399871474128, + "learning_rate": 3.6910203156786496e-05, + "loss": 0.102, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07761038094758987, + "step": 3320, + "valid_targets_mean": 1605.5, + "valid_targets_min": 781 + }, + { + "epoch": 1.832965821389195, + "grad_norm": 0.3418727177504921, + "learning_rate": 3.6895508479435964e-05, + "loss": 0.0987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04126843437552452, + "step": 3325, + "valid_targets_mean": 3645.4, + "valid_targets_min": 2141 + }, + { + "epoch": 1.8357221609702314, + "grad_norm": 0.4155077239287429, + "learning_rate": 3.688078188152639e-05, + "loss": 0.1008, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04582921415567398, + "step": 3330, + "valid_targets_mean": 1990.2, + "valid_targets_min": 545 + }, + { + "epoch": 1.8384785005512678, + "grad_norm": 0.3572053611869392, + "learning_rate": 3.6866023390880605e-05, + "loss": 0.1578, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05396253615617752, + "step": 3335, + "valid_targets_mean": 3322.1, + "valid_targets_min": 904 + }, + { + "epoch": 1.8412348401323042, + "grad_norm": 0.3521084079036859, + "learning_rate": 3.6851233035381746e-05, + "loss": 0.1325, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09199566394090652, + "step": 3340, + "valid_targets_mean": 3514.5, + "valid_targets_min": 2184 + }, + { + "epoch": 1.8439911797133406, + "grad_norm": 0.32312812073993197, + "learning_rate": 3.683641084297309e-05, + "loss": 0.1249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03621070832014084, + "step": 3345, + "valid_targets_mean": 1680.6, + "valid_targets_min": 595 + }, + { + "epoch": 1.846747519294377, + "grad_norm": 0.8950490394505807, + "learning_rate": 3.6821556841658116e-05, + "loss": 0.1437, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12488283962011337, + "step": 3350, + "valid_targets_mean": 1323.4, + "valid_targets_min": 575 + }, + { + "epoch": 1.8495038588754134, + "grad_norm": 0.6204132808862215, + "learning_rate": 3.680667105950037e-05, + "loss": 0.1677, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06576195359230042, + "step": 3355, + "valid_targets_mean": 1671.9, + "valid_targets_min": 691 + }, + { + "epoch": 1.8522601984564497, + "grad_norm": 0.49109788422263134, + "learning_rate": 3.6791753524623456e-05, + "loss": 0.1176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0662253275513649, + "step": 3360, + "valid_targets_mean": 1879.5, + "valid_targets_min": 687 + }, + { + "epoch": 1.8550165380374861, + "grad_norm": 0.3835126958931053, + "learning_rate": 3.6776804265210963e-05, + "loss": 0.1375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049352727830410004, + "step": 3365, + "valid_targets_mean": 3679.5, + "valid_targets_min": 2015 + }, + { + "epoch": 1.8577728776185225, + "grad_norm": 0.6469953047529032, + "learning_rate": 3.676182330950641e-05, + "loss": 0.1202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06020920351147652, + "step": 3370, + "valid_targets_mean": 1331.5, + "valid_targets_min": 520 + }, + { + "epoch": 1.8605292171995589, + "grad_norm": 0.29768318772495167, + "learning_rate": 3.67468106858132e-05, + "loss": 0.1103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03998548537492752, + "step": 3375, + "valid_targets_mean": 4111.8, + "valid_targets_min": 466 + }, + { + "epoch": 1.8632855567805953, + "grad_norm": 0.2802466848544559, + "learning_rate": 3.673176642249459e-05, + "loss": 0.1147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031226113438606262, + "step": 3380, + "valid_targets_mean": 5544.5, + "valid_targets_min": 4219 + }, + { + "epoch": 1.8660418963616316, + "grad_norm": 0.3090643579314698, + "learning_rate": 3.671669054797358e-05, + "loss": 0.1023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04054310917854309, + "step": 3385, + "valid_targets_mean": 3800.1, + "valid_targets_min": 456 + }, + { + "epoch": 1.868798235942668, + "grad_norm": 0.3099623060874446, + "learning_rate": 3.670158309073291e-05, + "loss": 0.101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04805191233754158, + "step": 3390, + "valid_targets_mean": 3487.1, + "valid_targets_min": 976 + }, + { + "epoch": 1.8715545755237044, + "grad_norm": 0.3710583717152331, + "learning_rate": 3.6686444079314995e-05, + "loss": 0.1127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0656682625412941, + "step": 3395, + "valid_targets_mean": 2431.6, + "valid_targets_min": 591 + }, + { + "epoch": 1.8743109151047408, + "grad_norm": 0.43501299633127366, + "learning_rate": 3.667127354232185e-05, + "loss": 0.12, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06742389500141144, + "step": 3400, + "valid_targets_mean": 3660.8, + "valid_targets_min": 2753 + }, + { + "epoch": 1.8770672546857772, + "grad_norm": 0.4096783751846429, + "learning_rate": 3.6656071508415064e-05, + "loss": 0.1207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05960790067911148, + "step": 3405, + "valid_targets_mean": 3661.6, + "valid_targets_min": 2892 + }, + { + "epoch": 1.8798235942668136, + "grad_norm": 0.3250236943893233, + "learning_rate": 3.6640838006315715e-05, + "loss": 0.1097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047024741768836975, + "step": 3410, + "valid_targets_mean": 3447.0, + "valid_targets_min": 1093 + }, + { + "epoch": 1.88257993384785, + "grad_norm": 0.41229380844379365, + "learning_rate": 3.6625573064804366e-05, + "loss": 0.1077, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06980597227811813, + "step": 3415, + "valid_targets_mean": 3798.8, + "valid_targets_min": 1861 + }, + { + "epoch": 1.8853362734288863, + "grad_norm": 0.34980546846598604, + "learning_rate": 3.661027671272094e-05, + "loss": 0.123, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045716214925050735, + "step": 3420, + "valid_targets_mean": 3005.0, + "valid_targets_min": 720 + }, + { + "epoch": 1.8880926130099227, + "grad_norm": 0.2931252741462535, + "learning_rate": 3.659494897896473e-05, + "loss": 0.0922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040638960897922516, + "step": 3425, + "valid_targets_mean": 3779.8, + "valid_targets_min": 2761 + }, + { + "epoch": 1.890848952590959, + "grad_norm": 0.45224939719129176, + "learning_rate": 3.65795898924943e-05, + "loss": 0.1234, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04979335144162178, + "step": 3430, + "valid_targets_mean": 2407.5, + "valid_targets_min": 588 + }, + { + "epoch": 1.8936052921719955, + "grad_norm": 0.3757092777776501, + "learning_rate": 3.656419948232747e-05, + "loss": 0.1162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05638939142227173, + "step": 3435, + "valid_targets_mean": 3408.8, + "valid_targets_min": 596 + }, + { + "epoch": 1.8963616317530319, + "grad_norm": 0.3891485793991898, + "learning_rate": 3.654877777754123e-05, + "loss": 0.1262, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06547284126281738, + "step": 3440, + "valid_targets_mean": 3530.0, + "valid_targets_min": 1636 + }, + { + "epoch": 1.8991179713340682, + "grad_norm": 0.39994425558736674, + "learning_rate": 3.653332480727168e-05, + "loss": 0.1186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06062936782836914, + "step": 3445, + "valid_targets_mean": 1950.8, + "valid_targets_min": 454 + }, + { + "epoch": 1.9018743109151046, + "grad_norm": 0.7301957965236198, + "learning_rate": 3.651784060071401e-05, + "loss": 0.1305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06483713537454605, + "step": 3450, + "valid_targets_mean": 819.5, + "valid_targets_min": 609 + }, + { + "epoch": 1.904630650496141, + "grad_norm": 0.5146238072148598, + "learning_rate": 3.6502325187122425e-05, + "loss": 0.1177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09546191990375519, + "step": 3455, + "valid_targets_mean": 2364.8, + "valid_targets_min": 893 + }, + { + "epoch": 1.9073869900771774, + "grad_norm": 0.3937348082669565, + "learning_rate": 3.648677859581006e-05, + "loss": 0.1231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04862872138619423, + "step": 3460, + "valid_targets_mean": 3029.5, + "valid_targets_min": 1140 + }, + { + "epoch": 1.9101433296582138, + "grad_norm": 0.41289581744527126, + "learning_rate": 3.6471200856148984e-05, + "loss": 0.1145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07206332683563232, + "step": 3465, + "valid_targets_mean": 3029.6, + "valid_targets_min": 1502 + }, + { + "epoch": 1.9128996692392501, + "grad_norm": 0.6087996832227536, + "learning_rate": 3.6455591997570115e-05, + "loss": 0.1059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06079067662358284, + "step": 3470, + "valid_targets_mean": 1295.8, + "valid_targets_min": 624 + }, + { + "epoch": 1.9156560088202865, + "grad_norm": 0.5338457437478052, + "learning_rate": 3.643995204956315e-05, + "loss": 0.1264, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07431945949792862, + "step": 3475, + "valid_targets_mean": 1481.6, + "valid_targets_min": 687 + }, + { + "epoch": 1.918412348401323, + "grad_norm": 0.3020298304731887, + "learning_rate": 3.6424281041676526e-05, + "loss": 0.1104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036015138030052185, + "step": 3480, + "valid_targets_mean": 3963.5, + "valid_targets_min": 3651 + }, + { + "epoch": 1.9211686879823593, + "grad_norm": 0.34245976814196183, + "learning_rate": 3.6408579003517347e-05, + "loss": 0.1003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06540755927562714, + "step": 3485, + "valid_targets_mean": 5041.0, + "valid_targets_min": 3813 + }, + { + "epoch": 1.9239250275633957, + "grad_norm": 0.31834126927827683, + "learning_rate": 3.639284596475138e-05, + "loss": 0.1102, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04796181991696358, + "step": 3490, + "valid_targets_mean": 2392.4, + "valid_targets_min": 757 + }, + { + "epoch": 1.926681367144432, + "grad_norm": 0.35905455546331694, + "learning_rate": 3.637708195510293e-05, + "loss": 0.0976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0564880445599556, + "step": 3495, + "valid_targets_mean": 3566.1, + "valid_targets_min": 664 + }, + { + "epoch": 1.9294377067254684, + "grad_norm": 0.3615558970400759, + "learning_rate": 3.636128700435481e-05, + "loss": 0.0937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0704079195857048, + "step": 3500, + "valid_targets_mean": 4990.0, + "valid_targets_min": 4088 + }, + { + "epoch": 1.9321940463065048, + "grad_norm": 0.4874630957837254, + "learning_rate": 3.634546114234833e-05, + "loss": 0.116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046812377870082855, + "step": 3505, + "valid_targets_mean": 3220.0, + "valid_targets_min": 779 + }, + { + "epoch": 1.9349503858875412, + "grad_norm": 0.4697853330163501, + "learning_rate": 3.632960439898315e-05, + "loss": 0.1289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05786087363958359, + "step": 3510, + "valid_targets_mean": 3469.2, + "valid_targets_min": 1084 + }, + { + "epoch": 1.9377067254685776, + "grad_norm": 0.35769815789631837, + "learning_rate": 3.631371680421732e-05, + "loss": 0.1102, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06483104825019836, + "step": 3515, + "valid_targets_mean": 3214.4, + "valid_targets_min": 700 + }, + { + "epoch": 1.940463065049614, + "grad_norm": 0.3094616980721087, + "learning_rate": 3.6297798388067126e-05, + "loss": 0.1092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04543289542198181, + "step": 3520, + "valid_targets_mean": 2815.9, + "valid_targets_min": 769 + }, + { + "epoch": 1.9432194046306503, + "grad_norm": 0.2976093014734913, + "learning_rate": 3.628184918060714e-05, + "loss": 0.1369, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047192711383104324, + "step": 3525, + "valid_targets_mean": 3541.8, + "valid_targets_min": 982 + }, + { + "epoch": 1.9459757442116867, + "grad_norm": 0.38433956984801587, + "learning_rate": 3.626586921197007e-05, + "loss": 0.1073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06252744793891907, + "step": 3530, + "valid_targets_mean": 3189.4, + "valid_targets_min": 1043 + }, + { + "epoch": 1.9487320837927231, + "grad_norm": 0.3757947723044546, + "learning_rate": 3.624985851234676e-05, + "loss": 0.1164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05645408481359482, + "step": 3535, + "valid_targets_mean": 3380.1, + "valid_targets_min": 1067 + }, + { + "epoch": 1.9514884233737595, + "grad_norm": 0.34494055376410554, + "learning_rate": 3.6233817111986096e-05, + "loss": 0.111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08223801106214523, + "step": 3540, + "valid_targets_mean": 3086.6, + "valid_targets_min": 901 + }, + { + "epoch": 1.9542447629547959, + "grad_norm": 0.5498850856052251, + "learning_rate": 3.621774504119498e-05, + "loss": 0.1606, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06803995370864868, + "step": 3545, + "valid_targets_mean": 1287.2, + "valid_targets_min": 567 + }, + { + "epoch": 1.9570011025358323, + "grad_norm": 0.4336667884671241, + "learning_rate": 3.620164233033826e-05, + "loss": 0.1251, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07722136378288269, + "step": 3550, + "valid_targets_mean": 3460.5, + "valid_targets_min": 831 + }, + { + "epoch": 1.9597574421168686, + "grad_norm": 0.40992459365792505, + "learning_rate": 3.618550900983867e-05, + "loss": 0.108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04521401971578598, + "step": 3555, + "valid_targets_mean": 2148.4, + "valid_targets_min": 632 + }, + { + "epoch": 1.962513781697905, + "grad_norm": 0.3740129757132556, + "learning_rate": 3.616934511017677e-05, + "loss": 0.109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05230919271707535, + "step": 3560, + "valid_targets_mean": 2711.9, + "valid_targets_min": 1583 + }, + { + "epoch": 1.9652701212789414, + "grad_norm": 0.3350647233268889, + "learning_rate": 3.615315066189089e-05, + "loss": 0.1063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046494483947753906, + "step": 3565, + "valid_targets_mean": 3267.4, + "valid_targets_min": 796 + }, + { + "epoch": 1.9680264608599778, + "grad_norm": 0.32948863626595565, + "learning_rate": 3.6136925695577085e-05, + "loss": 0.0999, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04918238893151283, + "step": 3570, + "valid_targets_mean": 3208.9, + "valid_targets_min": 952 + }, + { + "epoch": 1.9707828004410142, + "grad_norm": 0.4945769991771703, + "learning_rate": 3.612067024188907e-05, + "loss": 0.1388, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06730319559574127, + "step": 3575, + "valid_targets_mean": 2115.5, + "valid_targets_min": 885 + }, + { + "epoch": 1.9735391400220506, + "grad_norm": 0.38873888140595425, + "learning_rate": 3.6104384331538144e-05, + "loss": 0.0959, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05504176393151283, + "step": 3580, + "valid_targets_mean": 3298.5, + "valid_targets_min": 1018 + }, + { + "epoch": 1.976295479603087, + "grad_norm": 0.4373140191653495, + "learning_rate": 3.608806799529317e-05, + "loss": 0.1078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05517278611660004, + "step": 3585, + "valid_targets_mean": 2170.4, + "valid_targets_min": 782 + }, + { + "epoch": 1.9790518191841233, + "grad_norm": 0.42998879544439833, + "learning_rate": 3.607172126398046e-05, + "loss": 0.1092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06523957848548889, + "step": 3590, + "valid_targets_mean": 3039.9, + "valid_targets_min": 2543 + }, + { + "epoch": 1.98180815876516, + "grad_norm": 0.2792080229058037, + "learning_rate": 3.60553441684838e-05, + "loss": 0.1093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0526401624083519, + "step": 3595, + "valid_targets_mean": 3951.6, + "valid_targets_min": 2264 + }, + { + "epoch": 1.9845644983461963, + "grad_norm": 0.4607750214601947, + "learning_rate": 3.603893673974429e-05, + "loss": 0.0937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0675540417432785, + "step": 3600, + "valid_targets_mean": 4005.2, + "valid_targets_min": 2423 + }, + { + "epoch": 1.9873208379272327, + "grad_norm": 0.3441127552673337, + "learning_rate": 3.6022499008760374e-05, + "loss": 0.1059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06714865565299988, + "step": 3605, + "valid_targets_mean": 4350.4, + "valid_targets_min": 2545 + }, + { + "epoch": 1.990077177508269, + "grad_norm": 0.4999325556702379, + "learning_rate": 3.600603100658773e-05, + "loss": 0.1417, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09872396290302277, + "step": 3610, + "valid_targets_mean": 2167.4, + "valid_targets_min": 1523 + }, + { + "epoch": 1.9928335170893055, + "grad_norm": 0.299248236169364, + "learning_rate": 3.5989532764339254e-05, + "loss": 0.1618, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0317203588783741, + "step": 3615, + "valid_targets_mean": 3601.6, + "valid_targets_min": 1092 + }, + { + "epoch": 1.9955898566703418, + "grad_norm": 0.3154165912596279, + "learning_rate": 3.5973004313184923e-05, + "loss": 0.1083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040509629994630814, + "step": 3620, + "valid_targets_mean": 3158.4, + "valid_targets_min": 835 + }, + { + "epoch": 1.9983461962513782, + "grad_norm": 0.31414820972030505, + "learning_rate": 3.5956445684351837e-05, + "loss": 0.1044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04721730202436447, + "step": 3625, + "valid_targets_mean": 3701.5, + "valid_targets_min": 2543 + }, + { + "epoch": 2.0011025358324144, + "grad_norm": 0.49942786584669435, + "learning_rate": 3.5939856909124085e-05, + "loss": 0.1921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10459035634994507, + "step": 3630, + "valid_targets_mean": 8485.9, + "valid_targets_min": 6756 + }, + { + "epoch": 2.0038588754134508, + "grad_norm": 0.4236585672151837, + "learning_rate": 3.5923238018842715e-05, + "loss": 0.2349, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11589233577251434, + "step": 3635, + "valid_targets_mean": 6639.4, + "valid_targets_min": 5905 + }, + { + "epoch": 2.006615214994487, + "grad_norm": 0.46938828195311527, + "learning_rate": 3.590658904490568e-05, + "loss": 0.2233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10234090685844421, + "step": 3640, + "valid_targets_mean": 4087.4, + "valid_targets_min": 160 + }, + { + "epoch": 2.0093715545755235, + "grad_norm": 0.34159743909178264, + "learning_rate": 3.588991001876775e-05, + "loss": 0.2195, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09964543581008911, + "step": 3645, + "valid_targets_mean": 7917.6, + "valid_targets_min": 5567 + }, + { + "epoch": 2.01212789415656, + "grad_norm": 0.36531429814496014, + "learning_rate": 3.58732009719405e-05, + "loss": 0.2098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11228038370609283, + "step": 3650, + "valid_targets_mean": 7361.2, + "valid_targets_min": 5601 + }, + { + "epoch": 2.0148842337375963, + "grad_norm": 0.3700330596309308, + "learning_rate": 3.5856461935992194e-05, + "loss": 0.2134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10905633866786957, + "step": 3655, + "valid_targets_mean": 6850.4, + "valid_targets_min": 5085 + }, + { + "epoch": 2.0176405733186327, + "grad_norm": 0.4010501515432445, + "learning_rate": 3.583969294254779e-05, + "loss": 0.2297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1117079108953476, + "step": 3660, + "valid_targets_mean": 6890.8, + "valid_targets_min": 5057 + }, + { + "epoch": 2.020396912899669, + "grad_norm": 0.4048377596991482, + "learning_rate": 3.58228940232888e-05, + "loss": 0.2227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09861376881599426, + "step": 3665, + "valid_targets_mean": 6346.6, + "valid_targets_min": 3920 + }, + { + "epoch": 2.0231532524807054, + "grad_norm": 0.41503145372045425, + "learning_rate": 3.580606520995331e-05, + "loss": 0.2114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10600385069847107, + "step": 3670, + "valid_targets_mean": 6362.6, + "valid_targets_min": 4343 + }, + { + "epoch": 2.025909592061742, + "grad_norm": 0.5273436104896607, + "learning_rate": 3.578920653433588e-05, + "loss": 0.2051, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12005949020385742, + "step": 3675, + "valid_targets_mean": 5179.8, + "valid_targets_min": 136 + }, + { + "epoch": 2.028665931642778, + "grad_norm": 0.34831925323558494, + "learning_rate": 3.5772318028287464e-05, + "loss": 0.2088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09311504662036896, + "step": 3680, + "valid_targets_mean": 7313.2, + "valid_targets_min": 5219 + }, + { + "epoch": 2.0314222712238146, + "grad_norm": 0.34226988841322786, + "learning_rate": 3.5755399723715396e-05, + "loss": 0.2029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08901868760585785, + "step": 3685, + "valid_targets_mean": 7894.1, + "valid_targets_min": 5279 + }, + { + "epoch": 2.034178610804851, + "grad_norm": 0.38847739853222246, + "learning_rate": 3.573845165258331e-05, + "loss": 0.2132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10724466294050217, + "step": 3690, + "valid_targets_mean": 7899.9, + "valid_targets_min": 5983 + }, + { + "epoch": 2.0369349503858873, + "grad_norm": 0.3490561873157814, + "learning_rate": 3.572147384691107e-05, + "loss": 0.2021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09561130404472351, + "step": 3695, + "valid_targets_mean": 7612.1, + "valid_targets_min": 5224 + }, + { + "epoch": 2.0396912899669237, + "grad_norm": 0.3949502002085623, + "learning_rate": 3.5704466338774714e-05, + "loss": 0.2016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10251956433057785, + "step": 3700, + "valid_targets_mean": 7219.4, + "valid_targets_min": 5185 + }, + { + "epoch": 2.04244762954796, + "grad_norm": 0.37210740424721755, + "learning_rate": 3.5687429160306405e-05, + "loss": 0.2089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11447414755821228, + "step": 3705, + "valid_targets_mean": 7588.1, + "valid_targets_min": 5062 + }, + { + "epoch": 2.0452039691289965, + "grad_norm": 0.3554584796078063, + "learning_rate": 3.567036234369435e-05, + "loss": 0.2023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09706956148147583, + "step": 3710, + "valid_targets_mean": 7702.1, + "valid_targets_min": 5810 + }, + { + "epoch": 2.047960308710033, + "grad_norm": 0.36697619013666954, + "learning_rate": 3.565326592118277e-05, + "loss": 0.2048, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10394071787595749, + "step": 3715, + "valid_targets_mean": 7926.0, + "valid_targets_min": 5470 + }, + { + "epoch": 2.0507166482910693, + "grad_norm": 0.39067890653327714, + "learning_rate": 3.563613992507181e-05, + "loss": 0.2227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11315180361270905, + "step": 3720, + "valid_targets_mean": 7969.2, + "valid_targets_min": 5422 + }, + { + "epoch": 2.0534729878721056, + "grad_norm": 0.4808759271786463, + "learning_rate": 3.5618984387717475e-05, + "loss": 0.2266, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07919387519359589, + "step": 3725, + "valid_targets_mean": 2969.4, + "valid_targets_min": 1886 + }, + { + "epoch": 2.056229327453142, + "grad_norm": 0.3979350358966238, + "learning_rate": 3.5601799341531605e-05, + "loss": 0.1877, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08818945288658142, + "step": 3730, + "valid_targets_mean": 6684.6, + "valid_targets_min": 4885 + }, + { + "epoch": 2.0589856670341784, + "grad_norm": 0.4179624353021285, + "learning_rate": 3.558458481898178e-05, + "loss": 0.2016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11038625240325928, + "step": 3735, + "valid_targets_mean": 7565.5, + "valid_targets_min": 5456 + }, + { + "epoch": 2.061742006615215, + "grad_norm": 0.39425008559944125, + "learning_rate": 3.556734085259127e-05, + "loss": 0.2029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10945954918861389, + "step": 3740, + "valid_targets_mean": 7017.6, + "valid_targets_min": 5547 + }, + { + "epoch": 2.064498346196251, + "grad_norm": 0.388395426605411, + "learning_rate": 3.555006747493896e-05, + "loss": 0.2099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10158365219831467, + "step": 3745, + "valid_targets_mean": 7078.5, + "valid_targets_min": 4931 + }, + { + "epoch": 2.0672546857772875, + "grad_norm": 0.40522568692126465, + "learning_rate": 3.553276471865934e-05, + "loss": 0.2057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10279615223407745, + "step": 3750, + "valid_targets_mean": 6780.6, + "valid_targets_min": 5634 + }, + { + "epoch": 2.070011025358324, + "grad_norm": 0.3657429794747763, + "learning_rate": 3.551543261644236e-05, + "loss": 0.1982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09644211083650589, + "step": 3755, + "valid_targets_mean": 8042.4, + "valid_targets_min": 5492 + }, + { + "epoch": 2.0727673649393603, + "grad_norm": 0.3866861483564573, + "learning_rate": 3.549807120103343e-05, + "loss": 0.1788, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0847739726305008, + "step": 3760, + "valid_targets_mean": 6528.6, + "valid_targets_min": 3602 + }, + { + "epoch": 2.075523704520397, + "grad_norm": 0.3870312153441221, + "learning_rate": 3.548068050523336e-05, + "loss": 0.1835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09708192944526672, + "step": 3765, + "valid_targets_mean": 6736.2, + "valid_targets_min": 4945 + }, + { + "epoch": 2.078280044101433, + "grad_norm": 0.3968862556248612, + "learning_rate": 3.5463260561898246e-05, + "loss": 0.196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09872926771640778, + "step": 3770, + "valid_targets_mean": 6115.0, + "valid_targets_min": 4885 + }, + { + "epoch": 2.08103638368247, + "grad_norm": 0.5528327450003329, + "learning_rate": 3.544581140393947e-05, + "loss": 0.2023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13953599333763123, + "step": 3775, + "valid_targets_mean": 4565.0, + "valid_targets_min": 214 + }, + { + "epoch": 2.083792723263506, + "grad_norm": 0.3607535443853825, + "learning_rate": 3.54283330643236e-05, + "loss": 0.21, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10516291856765747, + "step": 3780, + "valid_targets_mean": 7283.1, + "valid_targets_min": 4916 + }, + { + "epoch": 2.0865490628445427, + "grad_norm": 0.41170755254287195, + "learning_rate": 3.541082557607231e-05, + "loss": 0.2227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10606412589550018, + "step": 3785, + "valid_targets_mean": 6411.9, + "valid_targets_min": 4488 + }, + { + "epoch": 2.0893054024255786, + "grad_norm": 0.41597600598084233, + "learning_rate": 3.53932889722624e-05, + "loss": 0.2158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11685401201248169, + "step": 3790, + "valid_targets_mean": 7681.2, + "valid_targets_min": 5767 + }, + { + "epoch": 2.0920617420066154, + "grad_norm": 0.35290493368669595, + "learning_rate": 3.5375723286025615e-05, + "loss": 0.2037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09948420524597168, + "step": 3795, + "valid_targets_mean": 7875.4, + "valid_targets_min": 5611 + }, + { + "epoch": 2.0948180815876514, + "grad_norm": 0.4039561708436965, + "learning_rate": 3.53581285505487e-05, + "loss": 0.2056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1129312664270401, + "step": 3800, + "valid_targets_mean": 7596.0, + "valid_targets_min": 5711 + }, + { + "epoch": 2.097574421168688, + "grad_norm": 0.3605931957151205, + "learning_rate": 3.534050479907324e-05, + "loss": 0.21, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11935107409954071, + "step": 3805, + "valid_targets_mean": 8694.8, + "valid_targets_min": 5112 + }, + { + "epoch": 2.1003307607497246, + "grad_norm": 0.3974985966278427, + "learning_rate": 3.532285206489567e-05, + "loss": 0.2057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11450357735157013, + "step": 3810, + "valid_targets_mean": 6940.6, + "valid_targets_min": 6063 + }, + { + "epoch": 2.103087100330761, + "grad_norm": 0.3496470471280533, + "learning_rate": 3.5305170381367165e-05, + "loss": 0.2096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09721643477678299, + "step": 3815, + "valid_targets_mean": 7110.2, + "valid_targets_min": 4811 + }, + { + "epoch": 2.1058434399117973, + "grad_norm": 0.38345125726725116, + "learning_rate": 3.528745978189361e-05, + "loss": 0.2081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09659512341022491, + "step": 3820, + "valid_targets_mean": 6677.2, + "valid_targets_min": 5076 + }, + { + "epoch": 2.1085997794928337, + "grad_norm": 0.4187847744032647, + "learning_rate": 3.52697202999355e-05, + "loss": 0.2113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10457277297973633, + "step": 3825, + "valid_targets_mean": 5899.2, + "valid_targets_min": 4676 + }, + { + "epoch": 2.11135611907387, + "grad_norm": 0.369889563380258, + "learning_rate": 3.52519519690079e-05, + "loss": 0.2032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09598414599895477, + "step": 3830, + "valid_targets_mean": 7114.4, + "valid_targets_min": 5248 + }, + { + "epoch": 2.1141124586549065, + "grad_norm": 0.6388602623923433, + "learning_rate": 3.5234154822680394e-05, + "loss": 0.1975, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06818205863237381, + "step": 3835, + "valid_targets_mean": 1548.5, + "valid_targets_min": 587 + }, + { + "epoch": 2.116868798235943, + "grad_norm": 0.43782117346482236, + "learning_rate": 3.5216328894577e-05, + "loss": 0.201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11772646754980087, + "step": 3840, + "valid_targets_mean": 7040.5, + "valid_targets_min": 4733 + }, + { + "epoch": 2.1196251378169793, + "grad_norm": 0.4294212994591944, + "learning_rate": 3.5198474218376106e-05, + "loss": 0.203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10188500583171844, + "step": 3845, + "valid_targets_mean": 6161.2, + "valid_targets_min": 5555 + }, + { + "epoch": 2.1223814773980156, + "grad_norm": 0.40598019877184627, + "learning_rate": 3.518059082781043e-05, + "loss": 0.2046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10654512047767639, + "step": 3850, + "valid_targets_mean": 6580.2, + "valid_targets_min": 4862 + }, + { + "epoch": 2.125137816979052, + "grad_norm": 0.39003394050660894, + "learning_rate": 3.516267875666692e-05, + "loss": 0.1997, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09445083141326904, + "step": 3855, + "valid_targets_mean": 6575.2, + "valid_targets_min": 3768 + }, + { + "epoch": 2.1278941565600884, + "grad_norm": 0.38887925748366947, + "learning_rate": 3.514473803878672e-05, + "loss": 0.2039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10348153114318848, + "step": 3860, + "valid_targets_mean": 7154.8, + "valid_targets_min": 4858 + }, + { + "epoch": 2.130650496141125, + "grad_norm": 0.40668902717952854, + "learning_rate": 3.51267687080651e-05, + "loss": 0.2065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10259948670864105, + "step": 3865, + "valid_targets_mean": 6462.0, + "valid_targets_min": 4583 + }, + { + "epoch": 2.133406835722161, + "grad_norm": 0.4122891385493265, + "learning_rate": 3.5108770798451376e-05, + "loss": 0.2097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10981473326683044, + "step": 3870, + "valid_targets_mean": 6308.5, + "valid_targets_min": 4766 + }, + { + "epoch": 2.1361631753031975, + "grad_norm": 0.4870616537407158, + "learning_rate": 3.5090744343948875e-05, + "loss": 0.1759, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11843350529670715, + "step": 3875, + "valid_targets_mean": 4534.4, + "valid_targets_min": 147 + }, + { + "epoch": 2.138919514884234, + "grad_norm": 0.3955534731041613, + "learning_rate": 3.507268937861484e-05, + "loss": 0.2041, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11233700811862946, + "step": 3880, + "valid_targets_mean": 6778.1, + "valid_targets_min": 5132 + }, + { + "epoch": 2.1416758544652703, + "grad_norm": 0.3859849236543228, + "learning_rate": 3.505460593656039e-05, + "loss": 0.1977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10187512636184692, + "step": 3885, + "valid_targets_mean": 6979.8, + "valid_targets_min": 4942 + }, + { + "epoch": 2.1444321940463067, + "grad_norm": 0.4152948203917163, + "learning_rate": 3.5036494051950414e-05, + "loss": 0.2052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09301692247390747, + "step": 3890, + "valid_targets_mean": 5762.2, + "valid_targets_min": 4777 + }, + { + "epoch": 2.147188533627343, + "grad_norm": 0.3876393126373711, + "learning_rate": 3.5018353759003586e-05, + "loss": 0.194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09162431210279465, + "step": 3895, + "valid_targets_mean": 6373.9, + "valid_targets_min": 4810 + }, + { + "epoch": 2.1499448732083795, + "grad_norm": 0.39070472490057856, + "learning_rate": 3.500018509199222e-05, + "loss": 0.2007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10435852408409119, + "step": 3900, + "valid_targets_mean": 6151.0, + "valid_targets_min": 5084 + }, + { + "epoch": 2.152701212789416, + "grad_norm": 0.3793329582153217, + "learning_rate": 3.4981988085242243e-05, + "loss": 0.2, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09225691854953766, + "step": 3905, + "valid_targets_mean": 6133.5, + "valid_targets_min": 5736 + }, + { + "epoch": 2.1554575523704522, + "grad_norm": 0.7469482177503595, + "learning_rate": 3.4963762773133126e-05, + "loss": 0.2151, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08406773209571838, + "step": 3910, + "valid_targets_mean": 1255.8, + "valid_targets_min": 824 + }, + { + "epoch": 2.1582138919514886, + "grad_norm": 0.7561866014878327, + "learning_rate": 3.494550919009782e-05, + "loss": 0.196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09892508387565613, + "step": 3915, + "valid_targets_mean": 1512.6, + "valid_targets_min": 959 + }, + { + "epoch": 2.160970231532525, + "grad_norm": 0.8010592736566573, + "learning_rate": 3.4927227370622675e-05, + "loss": 0.1952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10001963376998901, + "step": 3920, + "valid_targets_mean": 1302.9, + "valid_targets_min": 680 + }, + { + "epoch": 2.1637265711135614, + "grad_norm": 0.8340620315553992, + "learning_rate": 3.49089173492474e-05, + "loss": 0.1882, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08908869326114655, + "step": 3925, + "valid_targets_mean": 1212.4, + "valid_targets_min": 681 + }, + { + "epoch": 2.1664829106945978, + "grad_norm": 0.7717141645006369, + "learning_rate": 3.4890579160564985e-05, + "loss": 0.191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10738691687583923, + "step": 3930, + "valid_targets_mean": 1328.5, + "valid_targets_min": 698 + }, + { + "epoch": 2.169239250275634, + "grad_norm": 0.9226833518840986, + "learning_rate": 3.487221283922164e-05, + "loss": 0.1917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11393006145954132, + "step": 3935, + "valid_targets_mean": 1465.4, + "valid_targets_min": 820 + }, + { + "epoch": 2.1719955898566705, + "grad_norm": 0.79938167323959, + "learning_rate": 3.485381841991671e-05, + "loss": 0.1857, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07289096713066101, + "step": 3940, + "valid_targets_mean": 988.9, + "valid_targets_min": 574 + }, + { + "epoch": 2.174751929437707, + "grad_norm": 0.8021614700664188, + "learning_rate": 3.4835395937402636e-05, + "loss": 0.2069, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08854611217975616, + "step": 3945, + "valid_targets_mean": 1320.0, + "valid_targets_min": 946 + }, + { + "epoch": 2.1775082690187433, + "grad_norm": 0.7804518271317007, + "learning_rate": 3.481694542648488e-05, + "loss": 0.187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08625820279121399, + "step": 3950, + "valid_targets_mean": 1316.8, + "valid_targets_min": 606 + }, + { + "epoch": 2.1802646085997797, + "grad_norm": 0.741297925168656, + "learning_rate": 3.479846692202187e-05, + "loss": 0.196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05494421720504761, + "step": 3955, + "valid_targets_mean": 1086.2, + "valid_targets_min": 681 + }, + { + "epoch": 2.183020948180816, + "grad_norm": 0.7874466456379395, + "learning_rate": 3.4779960458924876e-05, + "loss": 0.1812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0940193310379982, + "step": 3960, + "valid_targets_mean": 1247.8, + "valid_targets_min": 964 + }, + { + "epoch": 2.1857772877618524, + "grad_norm": 0.8448128862355645, + "learning_rate": 3.476142607215805e-05, + "loss": 0.1976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09445527195930481, + "step": 3965, + "valid_targets_mean": 1373.6, + "valid_targets_min": 614 + }, + { + "epoch": 2.188533627342889, + "grad_norm": 0.8220903218316327, + "learning_rate": 3.474286379673826e-05, + "loss": 0.1849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0995856299996376, + "step": 3970, + "valid_targets_mean": 1265.9, + "valid_targets_min": 458 + }, + { + "epoch": 2.191289966923925, + "grad_norm": 0.8129207424640306, + "learning_rate": 3.472427366773508e-05, + "loss": 0.168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05568225309252739, + "step": 3975, + "valid_targets_mean": 890.8, + "valid_targets_min": 523 + }, + { + "epoch": 2.1940463065049616, + "grad_norm": 0.8781778332117279, + "learning_rate": 3.47056557202707e-05, + "loss": 0.1736, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08408491313457489, + "step": 3980, + "valid_targets_mean": 1058.6, + "valid_targets_min": 463 + }, + { + "epoch": 2.196802646085998, + "grad_norm": 0.731334165707712, + "learning_rate": 3.468700998951987e-05, + "loss": 0.1862, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08408807218074799, + "step": 3985, + "valid_targets_mean": 1290.1, + "valid_targets_min": 651 + }, + { + "epoch": 2.1995589856670343, + "grad_norm": 0.7984039752630898, + "learning_rate": 3.466833651070983e-05, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12665465474128723, + "step": 3990, + "valid_targets_mean": 1706.0, + "valid_targets_min": 480 + }, + { + "epoch": 2.2023153252480707, + "grad_norm": 0.8146461403594031, + "learning_rate": 3.464963531912024e-05, + "loss": 0.1923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07892197370529175, + "step": 3995, + "valid_targets_mean": 1091.4, + "valid_targets_min": 703 + }, + { + "epoch": 2.205071664829107, + "grad_norm": 0.7551309884001379, + "learning_rate": 3.463090645008313e-05, + "loss": 0.1805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05857735872268677, + "step": 4000, + "valid_targets_mean": 1131.1, + "valid_targets_min": 634 + }, + { + "epoch": 2.2078280044101435, + "grad_norm": 0.6926027091409118, + "learning_rate": 3.461214993898282e-05, + "loss": 0.1811, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08328820765018463, + "step": 4005, + "valid_targets_mean": 1196.1, + "valid_targets_min": 689 + }, + { + "epoch": 2.21058434399118, + "grad_norm": 0.7686404008918676, + "learning_rate": 3.459336582125584e-05, + "loss": 0.1947, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06859257817268372, + "step": 4010, + "valid_targets_mean": 1053.8, + "valid_targets_min": 611 + }, + { + "epoch": 2.2133406835722162, + "grad_norm": 0.7649565380332815, + "learning_rate": 3.4574554132390884e-05, + "loss": 0.1962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10310222208499908, + "step": 4015, + "valid_targets_mean": 1158.5, + "valid_targets_min": 709 + }, + { + "epoch": 2.2160970231532526, + "grad_norm": 0.7798787557183197, + "learning_rate": 3.455571490792873e-05, + "loss": 0.183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09928115457296371, + "step": 4020, + "valid_targets_mean": 1434.8, + "valid_targets_min": 939 + }, + { + "epoch": 2.218853362734289, + "grad_norm": 0.7569429567715914, + "learning_rate": 3.45368481834622e-05, + "loss": 0.1789, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09241408109664917, + "step": 4025, + "valid_targets_mean": 1342.6, + "valid_targets_min": 771 + }, + { + "epoch": 2.2216097023153254, + "grad_norm": 0.6974252644649428, + "learning_rate": 3.451795399463605e-05, + "loss": 0.1848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08954732865095139, + "step": 4030, + "valid_targets_mean": 1609.8, + "valid_targets_min": 731 + }, + { + "epoch": 2.224366041896362, + "grad_norm": 0.7955064427587922, + "learning_rate": 3.4499032377146926e-05, + "loss": 0.1864, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09537677466869354, + "step": 4035, + "valid_targets_mean": 1349.6, + "valid_targets_min": 760 + }, + { + "epoch": 2.227122381477398, + "grad_norm": 0.7411751128954694, + "learning_rate": 3.448008336674331e-05, + "loss": 0.1875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11113826185464859, + "step": 4040, + "valid_targets_mean": 1545.2, + "valid_targets_min": 681 + }, + { + "epoch": 2.2298787210584345, + "grad_norm": 0.7248548498933794, + "learning_rate": 3.446110699922541e-05, + "loss": 0.1806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08568568527698517, + "step": 4045, + "valid_targets_mean": 1202.4, + "valid_targets_min": 530 + }, + { + "epoch": 2.232635060639471, + "grad_norm": 0.7228369174557302, + "learning_rate": 3.4442103310445166e-05, + "loss": 0.1803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09517540037631989, + "step": 4050, + "valid_targets_mean": 1326.9, + "valid_targets_min": 644 + }, + { + "epoch": 2.2353914002205073, + "grad_norm": 0.7637876673195323, + "learning_rate": 3.442307233630608e-05, + "loss": 0.1846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1259092390537262, + "step": 4055, + "valid_targets_mean": 1571.0, + "valid_targets_min": 679 + }, + { + "epoch": 2.2381477398015437, + "grad_norm": 0.7716243559727414, + "learning_rate": 3.4404014112763236e-05, + "loss": 0.1852, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10674044489860535, + "step": 4060, + "valid_targets_mean": 1623.6, + "valid_targets_min": 553 + }, + { + "epoch": 2.24090407938258, + "grad_norm": 0.7213898339923301, + "learning_rate": 3.43849286758232e-05, + "loss": 0.1817, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09768370538949966, + "step": 4065, + "valid_targets_mean": 1366.5, + "valid_targets_min": 678 + }, + { + "epoch": 2.2436604189636165, + "grad_norm": 0.8114524263844314, + "learning_rate": 3.436581606154394e-05, + "loss": 0.1786, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09212210774421692, + "step": 4070, + "valid_targets_mean": 1135.9, + "valid_targets_min": 651 + }, + { + "epoch": 2.246416758544653, + "grad_norm": 0.6947218559232597, + "learning_rate": 3.434667630603478e-05, + "loss": 0.1746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07649809122085571, + "step": 4075, + "valid_targets_mean": 1333.0, + "valid_targets_min": 654 + }, + { + "epoch": 2.249173098125689, + "grad_norm": 0.7300211265981523, + "learning_rate": 3.4327509445456325e-05, + "loss": 0.1805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07454437017440796, + "step": 4080, + "valid_targets_mean": 1202.2, + "valid_targets_min": 596 + }, + { + "epoch": 2.2519294377067256, + "grad_norm": 0.7708860179884368, + "learning_rate": 3.430831551602038e-05, + "loss": 0.1805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08578019589185715, + "step": 4085, + "valid_targets_mean": 1081.6, + "valid_targets_min": 521 + }, + { + "epoch": 2.254685777287762, + "grad_norm": 0.7831450013883233, + "learning_rate": 3.4289094553989885e-05, + "loss": 0.174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10304030776023865, + "step": 4090, + "valid_targets_mean": 1253.1, + "valid_targets_min": 564 + }, + { + "epoch": 2.2574421168687984, + "grad_norm": 0.7788392738107885, + "learning_rate": 3.426984659567887e-05, + "loss": 0.1786, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08378520607948303, + "step": 4095, + "valid_targets_mean": 1132.0, + "valid_targets_min": 502 + }, + { + "epoch": 2.2601984564498347, + "grad_norm": 0.7796052843861695, + "learning_rate": 3.425057167745236e-05, + "loss": 0.1838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07891083508729935, + "step": 4100, + "valid_targets_mean": 1162.9, + "valid_targets_min": 566 + }, + { + "epoch": 2.262954796030871, + "grad_norm": 0.7323694190411288, + "learning_rate": 3.42312698357263e-05, + "loss": 0.1842, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06539064645767212, + "step": 4105, + "valid_targets_mean": 976.0, + "valid_targets_min": 487 + }, + { + "epoch": 2.2657111356119075, + "grad_norm": 0.6817268707440671, + "learning_rate": 3.421194110696754e-05, + "loss": 0.1819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10275379568338394, + "step": 4110, + "valid_targets_mean": 1648.9, + "valid_targets_min": 632 + }, + { + "epoch": 2.268467475192944, + "grad_norm": 0.6728512278119185, + "learning_rate": 3.419258552769369e-05, + "loss": 0.1843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08621636778116226, + "step": 4115, + "valid_targets_mean": 1437.2, + "valid_targets_min": 571 + }, + { + "epoch": 2.2712238147739803, + "grad_norm": 0.6740410715363998, + "learning_rate": 3.41732031344731e-05, + "loss": 0.1911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10089679807424545, + "step": 4120, + "valid_targets_mean": 1790.2, + "valid_targets_min": 773 + }, + { + "epoch": 2.2739801543550167, + "grad_norm": 0.696345715852232, + "learning_rate": 3.4153793963924794e-05, + "loss": 0.1852, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06655026227235794, + "step": 4125, + "valid_targets_mean": 1070.1, + "valid_targets_min": 571 + }, + { + "epoch": 2.276736493936053, + "grad_norm": 0.9309292898995856, + "learning_rate": 3.413435805271836e-05, + "loss": 0.1775, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08663757145404816, + "step": 4130, + "valid_targets_mean": 1284.4, + "valid_targets_min": 691 + }, + { + "epoch": 2.2794928335170894, + "grad_norm": 0.8331820854322315, + "learning_rate": 3.411489543757394e-05, + "loss": 0.1845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08867689967155457, + "step": 4135, + "valid_targets_mean": 1053.1, + "valid_targets_min": 591 + }, + { + "epoch": 2.282249173098126, + "grad_norm": 0.7292863773266242, + "learning_rate": 3.409540615526209e-05, + "loss": 0.1789, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09356434643268585, + "step": 4140, + "valid_targets_mean": 1155.0, + "valid_targets_min": 541 + }, + { + "epoch": 2.285005512679162, + "grad_norm": 0.7489558387888429, + "learning_rate": 3.407589024260378e-05, + "loss": 0.1708, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07981206476688385, + "step": 4145, + "valid_targets_mean": 1092.8, + "valid_targets_min": 674 + }, + { + "epoch": 2.2877618522601986, + "grad_norm": 0.834043900533965, + "learning_rate": 3.405634773647027e-05, + "loss": 0.1787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07238012552261353, + "step": 4150, + "valid_targets_mean": 1024.1, + "valid_targets_min": 649 + }, + { + "epoch": 2.290518191841235, + "grad_norm": 0.7484287784000153, + "learning_rate": 3.403677867378307e-05, + "loss": 0.1796, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07536542415618896, + "step": 4155, + "valid_targets_mean": 1059.4, + "valid_targets_min": 615 + }, + { + "epoch": 2.2932745314222713, + "grad_norm": 0.8116333738600883, + "learning_rate": 3.4017183091513884e-05, + "loss": 0.1824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11379533261060715, + "step": 4160, + "valid_targets_mean": 1500.2, + "valid_targets_min": 653 + }, + { + "epoch": 2.2960308710033077, + "grad_norm": 0.6717274276146304, + "learning_rate": 3.399756102668448e-05, + "loss": 0.18, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10971176624298096, + "step": 4165, + "valid_targets_mean": 1596.4, + "valid_targets_min": 489 + }, + { + "epoch": 2.298787210584344, + "grad_norm": 0.7609106795362893, + "learning_rate": 3.3977912516366685e-05, + "loss": 0.1773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09915685653686523, + "step": 4170, + "valid_targets_mean": 1408.9, + "valid_targets_min": 537 + }, + { + "epoch": 2.3015435501653805, + "grad_norm": 0.7948236902815039, + "learning_rate": 3.395823759768229e-05, + "loss": 0.1789, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13299033045768738, + "step": 4175, + "valid_targets_mean": 1640.2, + "valid_targets_min": 691 + }, + { + "epoch": 2.304299889746417, + "grad_norm": 0.7497133582823439, + "learning_rate": 3.393853630780297e-05, + "loss": 0.1791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12009297311306, + "step": 4180, + "valid_targets_mean": 1621.9, + "valid_targets_min": 1101 + }, + { + "epoch": 2.3070562293274532, + "grad_norm": 0.7402573543040295, + "learning_rate": 3.391880868395022e-05, + "loss": 0.1775, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12081624567508698, + "step": 4185, + "valid_targets_mean": 1632.5, + "valid_targets_min": 515 + }, + { + "epoch": 2.3098125689084896, + "grad_norm": 1.5147365391565963, + "learning_rate": 3.38990547633953e-05, + "loss": 0.1772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07442426681518555, + "step": 4190, + "valid_targets_mean": 1039.8, + "valid_targets_min": 672 + }, + { + "epoch": 2.312568908489526, + "grad_norm": 0.7975153556046493, + "learning_rate": 3.3879274583459145e-05, + "loss": 0.1724, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09240429103374481, + "step": 4195, + "valid_targets_mean": 1332.4, + "valid_targets_min": 746 + }, + { + "epoch": 2.3153252480705624, + "grad_norm": 0.794995010458824, + "learning_rate": 3.3859468181512305e-05, + "loss": 0.1841, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10990899801254272, + "step": 4200, + "valid_targets_mean": 1510.9, + "valid_targets_min": 684 + }, + { + "epoch": 2.3180815876515988, + "grad_norm": 0.7204980985683088, + "learning_rate": 3.3839635594974864e-05, + "loss": 0.1809, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08662702143192291, + "step": 4205, + "valid_targets_mean": 1377.9, + "valid_targets_min": 690 + }, + { + "epoch": 2.320837927232635, + "grad_norm": 0.7197067209944108, + "learning_rate": 3.381977686131639e-05, + "loss": 0.1729, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0923653095960617, + "step": 4210, + "valid_targets_mean": 1314.6, + "valid_targets_min": 983 + }, + { + "epoch": 2.3235942668136715, + "grad_norm": 0.7935914280565703, + "learning_rate": 3.379989201805582e-05, + "loss": 0.182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09884177148342133, + "step": 4215, + "valid_targets_mean": 1375.6, + "valid_targets_min": 875 + }, + { + "epoch": 2.326350606394708, + "grad_norm": 0.8176378023514344, + "learning_rate": 3.377998110276147e-05, + "loss": 0.1801, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10584904253482819, + "step": 4220, + "valid_targets_mean": 1341.9, + "valid_targets_min": 545 + }, + { + "epoch": 2.3291069459757443, + "grad_norm": 0.6976263423812986, + "learning_rate": 3.376004415305086e-05, + "loss": 0.1851, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07302886247634888, + "step": 4225, + "valid_targets_mean": 1162.1, + "valid_targets_min": 557 + }, + { + "epoch": 2.3318632855567807, + "grad_norm": 0.7757106695562364, + "learning_rate": 3.374008120659073e-05, + "loss": 0.1862, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0920407623052597, + "step": 4230, + "valid_targets_mean": 1242.5, + "valid_targets_min": 633 + }, + { + "epoch": 2.334619625137817, + "grad_norm": 0.7512866347623487, + "learning_rate": 3.372009230109694e-05, + "loss": 0.1802, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10986312478780746, + "step": 4235, + "valid_targets_mean": 1377.1, + "valid_targets_min": 507 + }, + { + "epoch": 2.3373759647188534, + "grad_norm": 0.8470554866500135, + "learning_rate": 3.3700077474334354e-05, + "loss": 0.1878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08202416449785233, + "step": 4240, + "valid_targets_mean": 972.0, + "valid_targets_min": 638 + }, + { + "epoch": 2.34013230429989, + "grad_norm": 0.8054967330390886, + "learning_rate": 3.3680036764116866e-05, + "loss": 0.1794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07533600926399231, + "step": 4245, + "valid_targets_mean": 1013.2, + "valid_targets_min": 508 + }, + { + "epoch": 2.342888643880926, + "grad_norm": 0.7439325724508226, + "learning_rate": 3.365997020830722e-05, + "loss": 0.1929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09420184791088104, + "step": 4250, + "valid_targets_mean": 1574.2, + "valid_targets_min": 1057 + }, + { + "epoch": 2.3456449834619626, + "grad_norm": 0.7450489191235881, + "learning_rate": 3.363987784481702e-05, + "loss": 0.186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06566415727138519, + "step": 4255, + "valid_targets_mean": 1055.1, + "valid_targets_min": 483 + }, + { + "epoch": 2.348401323042999, + "grad_norm": 0.7656976379446532, + "learning_rate": 3.361975971160662e-05, + "loss": 0.1861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08663272857666016, + "step": 4260, + "valid_targets_mean": 1155.2, + "valid_targets_min": 669 + }, + { + "epoch": 2.3511576626240354, + "grad_norm": 0.7463899410550351, + "learning_rate": 3.359961584668505e-05, + "loss": 0.181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11221029609441757, + "step": 4265, + "valid_targets_mean": 1534.2, + "valid_targets_min": 796 + }, + { + "epoch": 2.3539140022050717, + "grad_norm": 0.7694164008718813, + "learning_rate": 3.357944628810998e-05, + "loss": 0.1762, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.062227390706539154, + "step": 4270, + "valid_targets_mean": 822.2, + "valid_targets_min": 505 + }, + { + "epoch": 2.356670341786108, + "grad_norm": 0.7769627749244129, + "learning_rate": 3.35592510739876e-05, + "loss": 0.1868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06050511449575424, + "step": 4275, + "valid_targets_mean": 1010.2, + "valid_targets_min": 475 + }, + { + "epoch": 2.3594266813671445, + "grad_norm": 0.7121789569394993, + "learning_rate": 3.3539030242472574e-05, + "loss": 0.1783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09945737570524216, + "step": 4280, + "valid_targets_mean": 1546.2, + "valid_targets_min": 619 + }, + { + "epoch": 2.362183020948181, + "grad_norm": 0.7600904243910556, + "learning_rate": 3.351878383176797e-05, + "loss": 0.1738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10234350711107254, + "step": 4285, + "valid_targets_mean": 1300.4, + "valid_targets_min": 594 + }, + { + "epoch": 2.3649393605292173, + "grad_norm": 0.7588719711600311, + "learning_rate": 3.34985118801252e-05, + "loss": 0.1733, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06882726401090622, + "step": 4290, + "valid_targets_mean": 1015.0, + "valid_targets_min": 570 + }, + { + "epoch": 2.3676957001102537, + "grad_norm": 0.687865398332125, + "learning_rate": 3.347821442584389e-05, + "loss": 0.1916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10670061409473419, + "step": 4295, + "valid_targets_mean": 1735.2, + "valid_targets_min": 839 + }, + { + "epoch": 2.37045203969129, + "grad_norm": 0.7038202991776232, + "learning_rate": 3.3457891507271876e-05, + "loss": 0.1798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09059254825115204, + "step": 4300, + "valid_targets_mean": 1542.8, + "valid_targets_min": 851 + }, + { + "epoch": 2.3732083792723264, + "grad_norm": 0.7220139469203353, + "learning_rate": 3.34375431628051e-05, + "loss": 0.1765, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09545549750328064, + "step": 4305, + "valid_targets_mean": 1448.4, + "valid_targets_min": 818 + }, + { + "epoch": 2.375964718853363, + "grad_norm": 0.7907250536373589, + "learning_rate": 3.341716943088754e-05, + "loss": 0.1679, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08945708721876144, + "step": 4310, + "valid_targets_mean": 1311.9, + "valid_targets_min": 564 + }, + { + "epoch": 2.378721058434399, + "grad_norm": 0.7083948630737831, + "learning_rate": 3.3396770350011134e-05, + "loss": 0.1611, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08140061795711517, + "step": 4315, + "valid_targets_mean": 1382.9, + "valid_targets_min": 1083 + }, + { + "epoch": 2.3814773980154356, + "grad_norm": 0.8491577240238146, + "learning_rate": 3.3376345958715716e-05, + "loss": 0.1789, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10330396890640259, + "step": 4320, + "valid_targets_mean": 1268.8, + "valid_targets_min": 840 + }, + { + "epoch": 2.384233737596472, + "grad_norm": 0.7228077089229151, + "learning_rate": 3.335589629558894e-05, + "loss": 0.1789, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11730960756540298, + "step": 4325, + "valid_targets_mean": 1602.0, + "valid_targets_min": 639 + }, + { + "epoch": 2.3869900771775083, + "grad_norm": 0.8174316741387632, + "learning_rate": 3.33354213992662e-05, + "loss": 0.1753, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09316502511501312, + "step": 4330, + "valid_targets_mean": 1375.6, + "valid_targets_min": 655 + }, + { + "epoch": 2.3897464167585447, + "grad_norm": 0.6845318823322625, + "learning_rate": 3.3314921308430554e-05, + "loss": 0.1804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09287379682064056, + "step": 4335, + "valid_targets_mean": 1536.6, + "valid_targets_min": 736 + }, + { + "epoch": 2.392502756339581, + "grad_norm": 0.8418031949879158, + "learning_rate": 3.32943960618127e-05, + "loss": 0.1793, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10054989159107208, + "step": 4340, + "valid_targets_mean": 1363.6, + "valid_targets_min": 479 + }, + { + "epoch": 2.3952590959206175, + "grad_norm": 0.7139020854308563, + "learning_rate": 3.32738456981908e-05, + "loss": 0.1813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10856462270021439, + "step": 4345, + "valid_targets_mean": 2095.4, + "valid_targets_min": 1094 + }, + { + "epoch": 2.398015435501654, + "grad_norm": 0.6007331417961358, + "learning_rate": 3.325327025639054e-05, + "loss": 0.2882, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15076464414596558, + "step": 4350, + "valid_targets_mean": 3183.9, + "valid_targets_min": 343 + }, + { + "epoch": 2.4007717750826902, + "grad_norm": 0.5247796474050412, + "learning_rate": 3.3232669775284926e-05, + "loss": 0.2798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1524302214384079, + "step": 4355, + "valid_targets_mean": 4846.2, + "valid_targets_min": 1479 + }, + { + "epoch": 2.4035281146637266, + "grad_norm": 0.4785555187651168, + "learning_rate": 3.321204429379429e-05, + "loss": 0.2705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14491724967956543, + "step": 4360, + "valid_targets_mean": 5567.5, + "valid_targets_min": 2479 + }, + { + "epoch": 2.406284454244763, + "grad_norm": 0.44996462250237024, + "learning_rate": 3.3191393850886225e-05, + "loss": 0.2679, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13994614779949188, + "step": 4365, + "valid_targets_mean": 5882.1, + "valid_targets_min": 1354 + }, + { + "epoch": 2.4090407938257994, + "grad_norm": 0.5602885226825515, + "learning_rate": 3.317071848557544e-05, + "loss": 0.2749, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11292251944541931, + "step": 4370, + "valid_targets_mean": 2888.0, + "valid_targets_min": 830 + }, + { + "epoch": 2.4117971334068358, + "grad_norm": 0.5775633544217508, + "learning_rate": 3.315001823692376e-05, + "loss": 0.2802, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1343356966972351, + "step": 4375, + "valid_targets_mean": 3578.1, + "valid_targets_min": 1207 + }, + { + "epoch": 2.414553472987872, + "grad_norm": 0.47537552785301973, + "learning_rate": 3.3129293144039995e-05, + "loss": 0.2644, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13634787499904633, + "step": 4380, + "valid_targets_mean": 5160.0, + "valid_targets_min": 1843 + }, + { + "epoch": 2.4173098125689085, + "grad_norm": 0.46329372931536, + "learning_rate": 3.310854324607993e-05, + "loss": 0.2712, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11323797702789307, + "step": 4385, + "valid_targets_mean": 4569.2, + "valid_targets_min": 1212 + }, + { + "epoch": 2.420066152149945, + "grad_norm": 0.545922121014474, + "learning_rate": 3.308776858224619e-05, + "loss": 0.2797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15045878291130066, + "step": 4390, + "valid_targets_mean": 3779.9, + "valid_targets_min": 2023 + }, + { + "epoch": 2.4228224917309813, + "grad_norm": 0.5639616097370329, + "learning_rate": 3.3066969191788184e-05, + "loss": 0.2767, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1240997314453125, + "step": 4395, + "valid_targets_mean": 3853.4, + "valid_targets_min": 1365 + }, + { + "epoch": 2.4255788313120177, + "grad_norm": 0.5161370425225745, + "learning_rate": 3.304614511400205e-05, + "loss": 0.2783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14508754014968872, + "step": 4400, + "valid_targets_mean": 4372.9, + "valid_targets_min": 606 + }, + { + "epoch": 2.428335170893054, + "grad_norm": 0.5103076383690245, + "learning_rate": 3.3025296388230564e-05, + "loss": 0.2651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13774776458740234, + "step": 4405, + "valid_targets_mean": 4523.6, + "valid_targets_min": 502 + }, + { + "epoch": 2.4310915104740904, + "grad_norm": 0.8691548145605876, + "learning_rate": 3.300442305386308e-05, + "loss": 0.2735, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13216952979564667, + "step": 4410, + "valid_targets_mean": 4513.8, + "valid_targets_min": 1377 + }, + { + "epoch": 2.433847850055127, + "grad_norm": 0.7135368798461573, + "learning_rate": 3.2983525150335414e-05, + "loss": 0.267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16096174716949463, + "step": 4415, + "valid_targets_mean": 3668.9, + "valid_targets_min": 1587 + }, + { + "epoch": 2.436604189636163, + "grad_norm": 0.5969743332053989, + "learning_rate": 3.2962602717129834e-05, + "loss": 0.2794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15248140692710876, + "step": 4420, + "valid_targets_mean": 3505.9, + "valid_targets_min": 912 + }, + { + "epoch": 2.4393605292171996, + "grad_norm": 0.6671767326260614, + "learning_rate": 3.294165579377493e-05, + "loss": 0.2701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1224624365568161, + "step": 4425, + "valid_targets_mean": 3756.0, + "valid_targets_min": 2444 + }, + { + "epoch": 2.442116868798236, + "grad_norm": 0.5620294945112442, + "learning_rate": 3.292068441984558e-05, + "loss": 0.2698, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12923306226730347, + "step": 4430, + "valid_targets_mean": 3275.1, + "valid_targets_min": 600 + }, + { + "epoch": 2.4448732083792724, + "grad_norm": 0.6179373149197903, + "learning_rate": 3.2899688634962846e-05, + "loss": 0.2852, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1373014897108078, + "step": 4435, + "valid_targets_mean": 3077.1, + "valid_targets_min": 1475 + }, + { + "epoch": 2.4476295479603087, + "grad_norm": 0.5605532038452113, + "learning_rate": 3.287866847879389e-05, + "loss": 0.2778, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1274903565645218, + "step": 4440, + "valid_targets_mean": 3199.1, + "valid_targets_min": 853 + }, + { + "epoch": 2.450385887541345, + "grad_norm": 0.5692548986107292, + "learning_rate": 3.285762399105195e-05, + "loss": 0.273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13850440084934235, + "step": 4445, + "valid_targets_mean": 3320.5, + "valid_targets_min": 1256 + }, + { + "epoch": 2.4531422271223815, + "grad_norm": 0.555647542081339, + "learning_rate": 3.283655521149623e-05, + "loss": 0.2784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13061928749084473, + "step": 4450, + "valid_targets_mean": 3628.8, + "valid_targets_min": 1024 + }, + { + "epoch": 2.455898566703418, + "grad_norm": 0.5293555954973481, + "learning_rate": 3.281546217993181e-05, + "loss": 0.2673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11742659658193588, + "step": 4455, + "valid_targets_mean": 3396.1, + "valid_targets_min": 561 + }, + { + "epoch": 2.4586549062844543, + "grad_norm": 0.6325219145479504, + "learning_rate": 3.27943449362096e-05, + "loss": 0.2518, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11907361447811127, + "step": 4460, + "valid_targets_mean": 2552.5, + "valid_targets_min": 999 + }, + { + "epoch": 2.4614112458654906, + "grad_norm": 0.6021922041008461, + "learning_rate": 3.277320352022624e-05, + "loss": 0.2682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15030837059020996, + "step": 4465, + "valid_targets_mean": 3521.1, + "valid_targets_min": 1586 + }, + { + "epoch": 2.464167585446527, + "grad_norm": 0.6004063061050376, + "learning_rate": 3.275203797192406e-05, + "loss": 0.2755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11109170317649841, + "step": 4470, + "valid_targets_mean": 2330.2, + "valid_targets_min": 1008 + }, + { + "epoch": 2.4669239250275634, + "grad_norm": 0.4675422787415671, + "learning_rate": 3.2730848331290984e-05, + "loss": 0.265, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1115921288728714, + "step": 4475, + "valid_targets_mean": 3869.1, + "valid_targets_min": 924 + }, + { + "epoch": 2.4696802646086, + "grad_norm": 0.5352193530625332, + "learning_rate": 3.270963463836043e-05, + "loss": 0.2675, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16235104203224182, + "step": 4480, + "valid_targets_mean": 4625.8, + "valid_targets_min": 2490 + }, + { + "epoch": 2.472436604189636, + "grad_norm": 0.5768575559440492, + "learning_rate": 3.268839693321128e-05, + "loss": 0.2573, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16328275203704834, + "step": 4485, + "valid_targets_mean": 3477.4, + "valid_targets_min": 1270 + }, + { + "epoch": 2.4751929437706726, + "grad_norm": 0.6473028710403836, + "learning_rate": 3.266713525596776e-05, + "loss": 0.2754, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1547214239835739, + "step": 4490, + "valid_targets_mean": 3181.4, + "valid_targets_min": 935 + }, + { + "epoch": 2.477949283351709, + "grad_norm": 0.5323784972097392, + "learning_rate": 3.2645849646799406e-05, + "loss": 0.2585, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11820574104785919, + "step": 4495, + "valid_targets_mean": 3371.8, + "valid_targets_min": 1590 + }, + { + "epoch": 2.4807056229327453, + "grad_norm": 0.5481487469413873, + "learning_rate": 3.262454014592097e-05, + "loss": 0.2629, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14850397408008575, + "step": 4500, + "valid_targets_mean": 4045.0, + "valid_targets_min": 1500 + }, + { + "epoch": 2.4834619625137817, + "grad_norm": 0.6371244228269134, + "learning_rate": 3.260320679359232e-05, + "loss": 0.2621, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13558757305145264, + "step": 4505, + "valid_targets_mean": 2852.4, + "valid_targets_min": 1036 + }, + { + "epoch": 2.486218302094818, + "grad_norm": 0.6247999651309509, + "learning_rate": 3.258184963011842e-05, + "loss": 0.2648, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12605968117713928, + "step": 4510, + "valid_targets_mean": 3366.2, + "valid_targets_min": 994 + }, + { + "epoch": 2.4889746416758545, + "grad_norm": 0.6384686548813657, + "learning_rate": 3.2560468695849174e-05, + "loss": 0.2578, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12265051901340485, + "step": 4515, + "valid_targets_mean": 2559.0, + "valid_targets_min": 1178 + }, + { + "epoch": 2.491730981256891, + "grad_norm": 0.606645266217074, + "learning_rate": 3.253906403117945e-05, + "loss": 0.2728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14127217233181, + "step": 4520, + "valid_targets_mean": 3592.1, + "valid_targets_min": 1027 + }, + { + "epoch": 2.4944873208379272, + "grad_norm": 0.5572635008858075, + "learning_rate": 3.25176356765489e-05, + "loss": 0.269, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11955130845308304, + "step": 4525, + "valid_targets_mean": 3360.6, + "valid_targets_min": 1268 + }, + { + "epoch": 2.4972436604189636, + "grad_norm": 0.6058432293863946, + "learning_rate": 3.249618367244197e-05, + "loss": 0.2592, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11567891389131546, + "step": 4530, + "valid_targets_mean": 3154.4, + "valid_targets_min": 1092 + }, + { + "epoch": 2.5, + "grad_norm": 0.6199653931351148, + "learning_rate": 3.2474708059387775e-05, + "loss": 0.257, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13341371715068817, + "step": 4535, + "valid_targets_mean": 2666.2, + "valid_targets_min": 595 + }, + { + "epoch": 2.5027563395810364, + "grad_norm": 0.8544695797662304, + "learning_rate": 3.245320887796001e-05, + "loss": 0.2676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14412394165992737, + "step": 4540, + "valid_targets_mean": 3851.8, + "valid_targets_min": 1302 + }, + { + "epoch": 2.5055126791620728, + "grad_norm": 0.5953241729301344, + "learning_rate": 3.2431686168776944e-05, + "loss": 0.2609, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13021817803382874, + "step": 4545, + "valid_targets_mean": 2854.0, + "valid_targets_min": 1167 + }, + { + "epoch": 2.508269018743109, + "grad_norm": 0.777020855313702, + "learning_rate": 3.241013997250126e-05, + "loss": 0.2526, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1373911201953888, + "step": 4550, + "valid_targets_mean": 3929.6, + "valid_targets_min": 780 + }, + { + "epoch": 2.5110253583241455, + "grad_norm": 0.5569072264043813, + "learning_rate": 3.2388570329840025e-05, + "loss": 0.2718, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1268678605556488, + "step": 4555, + "valid_targets_mean": 3864.6, + "valid_targets_min": 473 + }, + { + "epoch": 2.513781697905182, + "grad_norm": 0.6146833868794787, + "learning_rate": 3.236697728154461e-05, + "loss": 0.2599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11360542476177216, + "step": 4560, + "valid_targets_mean": 2601.9, + "valid_targets_min": 801 + }, + { + "epoch": 2.5165380374862183, + "grad_norm": 0.5525255687962195, + "learning_rate": 3.234536086841061e-05, + "loss": 0.2552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14754009246826172, + "step": 4565, + "valid_targets_mean": 4882.4, + "valid_targets_min": 1807 + }, + { + "epoch": 2.5192943770672547, + "grad_norm": 0.6103215851223541, + "learning_rate": 3.232372113127774e-05, + "loss": 0.2476, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11828149110078812, + "step": 4570, + "valid_targets_mean": 2912.9, + "valid_targets_min": 1452 + }, + { + "epoch": 2.522050716648291, + "grad_norm": 0.5751023774014801, + "learning_rate": 3.230205811102979e-05, + "loss": 0.2756, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1362585425376892, + "step": 4575, + "valid_targets_mean": 3778.1, + "valid_targets_min": 1293 + }, + { + "epoch": 2.5248070562293274, + "grad_norm": 0.534170824165043, + "learning_rate": 3.228037184859456e-05, + "loss": 0.2528, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12485669553279877, + "step": 4580, + "valid_targets_mean": 3609.5, + "valid_targets_min": 657 + }, + { + "epoch": 2.527563395810364, + "grad_norm": 0.6164077020990025, + "learning_rate": 3.2258662384943745e-05, + "loss": 0.2734, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16018687188625336, + "step": 4585, + "valid_targets_mean": 3863.6, + "valid_targets_min": 2643 + }, + { + "epoch": 2.5303197353914, + "grad_norm": 0.5505146162984362, + "learning_rate": 3.2236929761092865e-05, + "loss": 0.2611, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1109701544046402, + "step": 4590, + "valid_targets_mean": 3228.8, + "valid_targets_min": 1043 + }, + { + "epoch": 2.5330760749724366, + "grad_norm": 0.5608885935980452, + "learning_rate": 3.221517401810121e-05, + "loss": 0.2689, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14458540081977844, + "step": 4595, + "valid_targets_mean": 3695.0, + "valid_targets_min": 1489 + }, + { + "epoch": 2.535832414553473, + "grad_norm": 0.6023904430968596, + "learning_rate": 3.219339519707173e-05, + "loss": 0.2518, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11732830852270126, + "step": 4600, + "valid_targets_mean": 2421.4, + "valid_targets_min": 617 + }, + { + "epoch": 2.5385887541345094, + "grad_norm": 0.5543473544923658, + "learning_rate": 3.217159333915101e-05, + "loss": 0.2732, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11485445499420166, + "step": 4605, + "valid_targets_mean": 3550.0, + "valid_targets_min": 1296 + }, + { + "epoch": 2.5413450937155457, + "grad_norm": 0.5787523711326675, + "learning_rate": 3.214976848552913e-05, + "loss": 0.2537, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16631287336349487, + "step": 4610, + "valid_targets_mean": 4005.0, + "valid_targets_min": 885 + }, + { + "epoch": 2.544101433296582, + "grad_norm": 0.5543542984347105, + "learning_rate": 3.212792067743963e-05, + "loss": 0.2473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1467181146144867, + "step": 4615, + "valid_targets_mean": 4549.0, + "valid_targets_min": 2263 + }, + { + "epoch": 2.5468577728776185, + "grad_norm": 0.5702847201481362, + "learning_rate": 3.21060499561594e-05, + "loss": 0.2408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12395124137401581, + "step": 4620, + "valid_targets_mean": 3004.1, + "valid_targets_min": 803 + }, + { + "epoch": 2.549614112458655, + "grad_norm": 0.5662943223519433, + "learning_rate": 3.208415636300864e-05, + "loss": 0.271, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12245997041463852, + "step": 4625, + "valid_targets_mean": 3451.6, + "valid_targets_min": 1962 + }, + { + "epoch": 2.5523704520396913, + "grad_norm": 0.5528487351271515, + "learning_rate": 3.2062239939350756e-05, + "loss": 0.2599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12461811304092407, + "step": 4630, + "valid_targets_mean": 3600.5, + "valid_targets_min": 754 + }, + { + "epoch": 2.5551267916207276, + "grad_norm": 0.5297770208582013, + "learning_rate": 3.204030072659229e-05, + "loss": 0.2624, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14221782982349396, + "step": 4635, + "valid_targets_mean": 4985.2, + "valid_targets_min": 1189 + }, + { + "epoch": 2.557883131201764, + "grad_norm": 0.5855360698154852, + "learning_rate": 3.201833876618281e-05, + "loss": 0.2583, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13874652981758118, + "step": 4640, + "valid_targets_mean": 3769.6, + "valid_targets_min": 2492 + }, + { + "epoch": 2.5606394707828004, + "grad_norm": 0.588467476019831, + "learning_rate": 3.199635409961491e-05, + "loss": 0.2677, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16264459490776062, + "step": 4645, + "valid_targets_mean": 4334.0, + "valid_targets_min": 1209 + }, + { + "epoch": 2.563395810363837, + "grad_norm": 0.5943313644271305, + "learning_rate": 3.197434676842404e-05, + "loss": 0.2623, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11739221960306168, + "step": 4650, + "valid_targets_mean": 2842.6, + "valid_targets_min": 798 + }, + { + "epoch": 2.566152149944873, + "grad_norm": 0.5245480872080829, + "learning_rate": 3.19523168141885e-05, + "loss": 0.2526, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1317036747932434, + "step": 4655, + "valid_targets_mean": 4873.6, + "valid_targets_min": 1525 + }, + { + "epoch": 2.5689084895259096, + "grad_norm": 0.5345268002640416, + "learning_rate": 3.193026427852932e-05, + "loss": 0.2503, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1186186671257019, + "step": 4660, + "valid_targets_mean": 3656.6, + "valid_targets_min": 1321 + }, + { + "epoch": 2.571664829106946, + "grad_norm": 0.6303397281899912, + "learning_rate": 3.190818920311018e-05, + "loss": 0.2489, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15391257405281067, + "step": 4665, + "valid_targets_mean": 4593.5, + "valid_targets_min": 1345 + }, + { + "epoch": 2.5744211686879823, + "grad_norm": 0.6781511131868172, + "learning_rate": 3.1886091629637376e-05, + "loss": 0.2502, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1218167394399643, + "step": 4670, + "valid_targets_mean": 2703.9, + "valid_targets_min": 1116 + }, + { + "epoch": 2.5771775082690187, + "grad_norm": 0.5881869171548327, + "learning_rate": 3.186397159985967e-05, + "loss": 0.2424, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14814725518226624, + "step": 4675, + "valid_targets_mean": 4335.2, + "valid_targets_min": 1280 + }, + { + "epoch": 2.579933847850055, + "grad_norm": 0.5798054072991303, + "learning_rate": 3.1841829155568284e-05, + "loss": 0.2463, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1119011715054512, + "step": 4680, + "valid_targets_mean": 3192.5, + "valid_targets_min": 967 + }, + { + "epoch": 2.5826901874310915, + "grad_norm": 0.5506353126353801, + "learning_rate": 3.181966433859676e-05, + "loss": 0.2473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1588534116744995, + "step": 4685, + "valid_targets_mean": 4250.6, + "valid_targets_min": 1640 + }, + { + "epoch": 2.585446527012128, + "grad_norm": 0.6822496164525904, + "learning_rate": 3.179747719082094e-05, + "loss": 0.2676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15511974692344666, + "step": 4690, + "valid_targets_mean": 2923.2, + "valid_targets_min": 530 + }, + { + "epoch": 2.5882028665931642, + "grad_norm": 0.5259242939437982, + "learning_rate": 3.1775267754158824e-05, + "loss": 0.2598, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1146039292216301, + "step": 4695, + "valid_targets_mean": 4365.1, + "valid_targets_min": 2153 + }, + { + "epoch": 2.5909592061742006, + "grad_norm": 0.6414234996780864, + "learning_rate": 3.1753036070570554e-05, + "loss": 0.2556, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15841686725616455, + "step": 4700, + "valid_targets_mean": 3581.8, + "valid_targets_min": 941 + }, + { + "epoch": 2.593715545755237, + "grad_norm": 0.548533110805486, + "learning_rate": 3.1730782182058255e-05, + "loss": 0.2518, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.128389373421669, + "step": 4705, + "valid_targets_mean": 4051.1, + "valid_targets_min": 1604 + }, + { + "epoch": 2.5964718853362734, + "grad_norm": 0.5776347816014533, + "learning_rate": 3.170850613066607e-05, + "loss": 0.2455, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13717395067214966, + "step": 4710, + "valid_targets_mean": 3441.6, + "valid_targets_min": 1242 + }, + { + "epoch": 2.5992282249173098, + "grad_norm": 0.5159615314895661, + "learning_rate": 3.168620795847994e-05, + "loss": 0.2448, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10139738023281097, + "step": 4715, + "valid_targets_mean": 3012.2, + "valid_targets_min": 990 + }, + { + "epoch": 2.601984564498346, + "grad_norm": 0.5578893816348409, + "learning_rate": 3.1663887707627666e-05, + "loss": 0.2485, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1140056625008583, + "step": 4720, + "valid_targets_mean": 4027.6, + "valid_targets_min": 947 + }, + { + "epoch": 2.6047409040793825, + "grad_norm": 0.598371208825273, + "learning_rate": 3.164154542027872e-05, + "loss": 0.2497, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11834971606731415, + "step": 4725, + "valid_targets_mean": 3011.8, + "valid_targets_min": 1424 + }, + { + "epoch": 2.607497243660419, + "grad_norm": 0.5548259520340277, + "learning_rate": 3.161918113864423e-05, + "loss": 0.2516, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14981065690517426, + "step": 4730, + "valid_targets_mean": 4182.6, + "valid_targets_min": 1308 + }, + { + "epoch": 2.6102535832414553, + "grad_norm": 0.5671658900757199, + "learning_rate": 3.159679490497686e-05, + "loss": 0.245, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08791469037532806, + "step": 4735, + "valid_targets_mean": 2722.2, + "valid_targets_min": 1081 + }, + { + "epoch": 2.6130099228224917, + "grad_norm": 0.5957049526028998, + "learning_rate": 3.157438676157075e-05, + "loss": 0.2668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11219050735235214, + "step": 4740, + "valid_targets_mean": 2829.9, + "valid_targets_min": 1961 + }, + { + "epoch": 2.615766262403528, + "grad_norm": 0.5413191320338946, + "learning_rate": 3.1551956750761444e-05, + "loss": 0.2624, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15696237981319427, + "step": 4745, + "valid_targets_mean": 4530.9, + "valid_targets_min": 827 + }, + { + "epoch": 2.6185226019845644, + "grad_norm": 0.5869527869681508, + "learning_rate": 3.15295049149258e-05, + "loss": 0.259, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14076007902622223, + "step": 4750, + "valid_targets_mean": 3167.5, + "valid_targets_min": 1162 + }, + { + "epoch": 2.621278941565601, + "grad_norm": 0.5641810077898917, + "learning_rate": 3.15070312964819e-05, + "loss": 0.267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1087508499622345, + "step": 4755, + "valid_targets_mean": 3328.9, + "valid_targets_min": 545 + }, + { + "epoch": 2.624035281146637, + "grad_norm": 0.5759538731099506, + "learning_rate": 3.148453593788899e-05, + "loss": 0.2577, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13121068477630615, + "step": 4760, + "valid_targets_mean": 3686.2, + "valid_targets_min": 1418 + }, + { + "epoch": 2.6267916207276736, + "grad_norm": 0.5514482238941472, + "learning_rate": 3.146201888164738e-05, + "loss": 0.2641, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12384310364723206, + "step": 4765, + "valid_targets_mean": 3815.1, + "valid_targets_min": 935 + }, + { + "epoch": 2.62954796030871, + "grad_norm": 0.6241199740990406, + "learning_rate": 3.143948017029838e-05, + "loss": 0.2648, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12936490774154663, + "step": 4770, + "valid_targets_mean": 2886.5, + "valid_targets_min": 755 + }, + { + "epoch": 2.6323042998897463, + "grad_norm": 0.5549709596894824, + "learning_rate": 3.141691984642421e-05, + "loss": 0.2525, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08658662438392639, + "step": 4775, + "valid_targets_mean": 2600.0, + "valid_targets_min": 306 + }, + { + "epoch": 2.6350606394707827, + "grad_norm": 0.5891216899813081, + "learning_rate": 3.139433795264791e-05, + "loss": 0.2571, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15226268768310547, + "step": 4780, + "valid_targets_mean": 3270.1, + "valid_targets_min": 753 + }, + { + "epoch": 2.637816979051819, + "grad_norm": 0.6368879365767073, + "learning_rate": 3.137173453163332e-05, + "loss": 0.2564, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12102415412664413, + "step": 4785, + "valid_targets_mean": 2390.8, + "valid_targets_min": 745 + }, + { + "epoch": 2.6405733186328555, + "grad_norm": 0.6057032668416162, + "learning_rate": 3.134910962608487e-05, + "loss": 0.262, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12719587981700897, + "step": 4790, + "valid_targets_mean": 3068.8, + "valid_targets_min": 365 + }, + { + "epoch": 2.643329658213892, + "grad_norm": 0.5535669778969304, + "learning_rate": 3.132646327874766e-05, + "loss": 0.2575, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10827545821666718, + "step": 4795, + "valid_targets_mean": 3204.9, + "valid_targets_min": 1726 + }, + { + "epoch": 2.6460859977949283, + "grad_norm": 0.5170691077917884, + "learning_rate": 3.130379553240726e-05, + "loss": 0.2376, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14567619562149048, + "step": 4800, + "valid_targets_mean": 5198.5, + "valid_targets_min": 2265 + }, + { + "epoch": 2.6488423373759646, + "grad_norm": 0.624860282120264, + "learning_rate": 3.1281106429889655e-05, + "loss": 0.2531, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11521455645561218, + "step": 4805, + "valid_targets_mean": 2861.9, + "valid_targets_min": 738 + }, + { + "epoch": 2.651598676957001, + "grad_norm": 0.5867940979712827, + "learning_rate": 3.125839601406123e-05, + "loss": 0.2393, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1153314858675003, + "step": 4810, + "valid_targets_mean": 2756.8, + "valid_targets_min": 864 + }, + { + "epoch": 2.6543550165380374, + "grad_norm": 0.5688029902011791, + "learning_rate": 3.123566432782858e-05, + "loss": 0.2491, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09928299486637115, + "step": 4815, + "valid_targets_mean": 2844.2, + "valid_targets_min": 806 + }, + { + "epoch": 2.657111356119074, + "grad_norm": 0.588558869054178, + "learning_rate": 3.121291141413853e-05, + "loss": 0.2523, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13748595118522644, + "step": 4820, + "valid_targets_mean": 4167.8, + "valid_targets_min": 2064 + }, + { + "epoch": 2.65986769570011, + "grad_norm": 0.5330056287141938, + "learning_rate": 3.1190137315977976e-05, + "loss": 0.2582, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13090115785598755, + "step": 4825, + "valid_targets_mean": 4761.1, + "valid_targets_min": 2688 + }, + { + "epoch": 2.6626240352811466, + "grad_norm": 0.6273116213966831, + "learning_rate": 3.116734207637386e-05, + "loss": 0.2443, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15392455458641052, + "step": 4830, + "valid_targets_mean": 3670.4, + "valid_targets_min": 1341 + }, + { + "epoch": 2.665380374862183, + "grad_norm": 0.627117260763579, + "learning_rate": 3.114452573839306e-05, + "loss": 0.2548, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14053672552108765, + "step": 4835, + "valid_targets_mean": 3385.0, + "valid_targets_min": 1728 + }, + { + "epoch": 2.6681367144432193, + "grad_norm": 0.5862567101772884, + "learning_rate": 3.112168834514231e-05, + "loss": 0.2536, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14603927731513977, + "step": 4840, + "valid_targets_mean": 3609.8, + "valid_targets_min": 899 + }, + { + "epoch": 2.6708930540242557, + "grad_norm": 0.5575365057365534, + "learning_rate": 3.1098829939768134e-05, + "loss": 0.2413, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12288836389780045, + "step": 4845, + "valid_targets_mean": 3314.1, + "valid_targets_min": 1395 + }, + { + "epoch": 2.673649393605292, + "grad_norm": 0.6084701640910142, + "learning_rate": 3.107595056545675e-05, + "loss": 0.2359, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1183367371559143, + "step": 4850, + "valid_targets_mean": 3130.0, + "valid_targets_min": 1190 + }, + { + "epoch": 2.6764057331863285, + "grad_norm": 0.5574086747952774, + "learning_rate": 3.1053050265433975e-05, + "loss": 0.2442, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13227033615112305, + "step": 4855, + "valid_targets_mean": 4297.4, + "valid_targets_min": 2075 + }, + { + "epoch": 2.679162072767365, + "grad_norm": 0.5684476639119265, + "learning_rate": 3.103012908296519e-05, + "loss": 0.2421, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14949369430541992, + "step": 4860, + "valid_targets_mean": 3947.9, + "valid_targets_min": 1983 + }, + { + "epoch": 2.6819184123484012, + "grad_norm": 0.5261137012856231, + "learning_rate": 3.100718706135521e-05, + "loss": 0.2312, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09753028303384781, + "step": 4865, + "valid_targets_mean": 3703.5, + "valid_targets_min": 841 + }, + { + "epoch": 2.6846747519294376, + "grad_norm": 0.5711361527371527, + "learning_rate": 3.098422424394824e-05, + "loss": 0.2365, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1031433492898941, + "step": 4870, + "valid_targets_mean": 3100.9, + "valid_targets_min": 1101 + }, + { + "epoch": 2.687431091510474, + "grad_norm": 0.5368546110392993, + "learning_rate": 3.096124067412774e-05, + "loss": 0.2432, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13342970609664917, + "step": 4875, + "valid_targets_mean": 3936.2, + "valid_targets_min": 1181 + }, + { + "epoch": 2.6901874310915104, + "grad_norm": 0.5384951652778744, + "learning_rate": 3.093823639531641e-05, + "loss": 0.2414, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09133481979370117, + "step": 4880, + "valid_targets_mean": 3304.5, + "valid_targets_min": 1244 + }, + { + "epoch": 2.6929437706725468, + "grad_norm": 0.6384228707730213, + "learning_rate": 3.091521145097606e-05, + "loss": 0.2714, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13877201080322266, + "step": 4885, + "valid_targets_mean": 2928.8, + "valid_targets_min": 652 + }, + { + "epoch": 2.695700110253583, + "grad_norm": 0.5502258399652725, + "learning_rate": 3.0892165884607545e-05, + "loss": 0.2392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13432471454143524, + "step": 4890, + "valid_targets_mean": 4858.0, + "valid_targets_min": 2492 + }, + { + "epoch": 2.6984564498346195, + "grad_norm": 0.5968454335576021, + "learning_rate": 3.086909973975069e-05, + "loss": 0.2437, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13325020670890808, + "step": 4895, + "valid_targets_mean": 3879.1, + "valid_targets_min": 1748 + }, + { + "epoch": 2.701212789415656, + "grad_norm": 0.5746184454730913, + "learning_rate": 3.084601305998419e-05, + "loss": 0.2402, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11013787984848022, + "step": 4900, + "valid_targets_mean": 3280.5, + "valid_targets_min": 2151 + }, + { + "epoch": 2.7039691289966923, + "grad_norm": 0.667006113823302, + "learning_rate": 3.082290588892553e-05, + "loss": 0.2505, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12861640751361847, + "step": 4905, + "valid_targets_mean": 3063.5, + "valid_targets_min": 1395 + }, + { + "epoch": 2.7067254685777287, + "grad_norm": 0.604621714262756, + "learning_rate": 3.079977827023092e-05, + "loss": 0.2529, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10132194310426712, + "step": 4910, + "valid_targets_mean": 2449.1, + "valid_targets_min": 576 + }, + { + "epoch": 2.709481808158765, + "grad_norm": 0.5257959275465992, + "learning_rate": 3.077663024759521e-05, + "loss": 0.2502, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11264118552207947, + "step": 4915, + "valid_targets_mean": 3442.4, + "valid_targets_min": 1108 + }, + { + "epoch": 2.7122381477398014, + "grad_norm": 0.5718323562212394, + "learning_rate": 3.075346186475178e-05, + "loss": 0.2448, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11069930344820023, + "step": 4920, + "valid_targets_mean": 3350.8, + "valid_targets_min": 759 + }, + { + "epoch": 2.714994487320838, + "grad_norm": 0.6334989632744245, + "learning_rate": 3.0730273165472486e-05, + "loss": 0.2463, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11604931950569153, + "step": 4925, + "valid_targets_mean": 2877.4, + "valid_targets_min": 2089 + }, + { + "epoch": 2.717750826901874, + "grad_norm": 0.609391097957824, + "learning_rate": 3.070706419356756e-05, + "loss": 0.2479, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12681584060192108, + "step": 4930, + "valid_targets_mean": 3304.8, + "valid_targets_min": 1116 + }, + { + "epoch": 2.7205071664829106, + "grad_norm": 0.4941092382575897, + "learning_rate": 3.0683834992885535e-05, + "loss": 0.1728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07402769476175308, + "step": 4935, + "valid_targets_mean": 2719.9, + "valid_targets_min": 540 + }, + { + "epoch": 2.723263506063947, + "grad_norm": 0.36426956481853273, + "learning_rate": 3.066058560731319e-05, + "loss": 0.1046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04080020636320114, + "step": 4940, + "valid_targets_mean": 3632.6, + "valid_targets_min": 2726 + }, + { + "epoch": 2.7260198456449833, + "grad_norm": 0.3714774691974349, + "learning_rate": 3.0637316080775396e-05, + "loss": 0.099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03908015787601471, + "step": 4945, + "valid_targets_mean": 3739.5, + "valid_targets_min": 2923 + }, + { + "epoch": 2.7287761852260197, + "grad_norm": 0.3704325636102285, + "learning_rate": 3.0614026457235114e-05, + "loss": 0.108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053242288529872894, + "step": 4950, + "valid_targets_mean": 3528.4, + "valid_targets_min": 2626 + }, + { + "epoch": 2.731532524807056, + "grad_norm": 0.43409878080877556, + "learning_rate": 3.059071678069324e-05, + "loss": 0.0958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055281296372413635, + "step": 4955, + "valid_targets_mean": 2684.9, + "valid_targets_min": 557 + }, + { + "epoch": 2.7342888643880925, + "grad_norm": 0.3891639177996408, + "learning_rate": 3.0567387095188586e-05, + "loss": 0.1028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050235457718372345, + "step": 4960, + "valid_targets_mean": 2843.5, + "valid_targets_min": 997 + }, + { + "epoch": 2.737045203969129, + "grad_norm": 0.6557972552479208, + "learning_rate": 3.054403744479774e-05, + "loss": 0.163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09889250993728638, + "step": 4965, + "valid_targets_mean": 1788.0, + "valid_targets_min": 920 + }, + { + "epoch": 2.7398015435501653, + "grad_norm": 0.3647379826635409, + "learning_rate": 3.0520667873635044e-05, + "loss": 0.1009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044200267642736435, + "step": 4970, + "valid_targets_mean": 3279.5, + "valid_targets_min": 1032 + }, + { + "epoch": 2.7425578831312016, + "grad_norm": 0.5479702259466633, + "learning_rate": 3.0497278425852445e-05, + "loss": 0.124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10118675231933594, + "step": 4975, + "valid_targets_mean": 2061.0, + "valid_targets_min": 878 + }, + { + "epoch": 2.745314222712238, + "grad_norm": 0.22616148005704528, + "learning_rate": 3.0473869145639453e-05, + "loss": 0.1196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030228998512029648, + "step": 4980, + "valid_targets_mean": 5122.1, + "valid_targets_min": 3479 + }, + { + "epoch": 2.7480705622932744, + "grad_norm": 0.3282941052688967, + "learning_rate": 3.0450440077223037e-05, + "loss": 0.0924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03108176589012146, + "step": 4985, + "valid_targets_mean": 3110.0, + "valid_targets_min": 760 + }, + { + "epoch": 2.750826901874311, + "grad_norm": 0.34447911187071256, + "learning_rate": 3.0426991264867586e-05, + "loss": 0.0985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04258149862289429, + "step": 4990, + "valid_targets_mean": 3693.5, + "valid_targets_min": 768 + }, + { + "epoch": 2.753583241455347, + "grad_norm": 0.4302125961624912, + "learning_rate": 3.0403522752874754e-05, + "loss": 0.0978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049686383455991745, + "step": 4995, + "valid_targets_mean": 2467.9, + "valid_targets_min": 489 + }, + { + "epoch": 2.7563395810363835, + "grad_norm": 0.2889296448338924, + "learning_rate": 3.0380034585583425e-05, + "loss": 0.1415, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03915070742368698, + "step": 5000, + "valid_targets_mean": 4141.6, + "valid_targets_min": 1952 + }, + { + "epoch": 2.75909592061742, + "grad_norm": 0.2818994107742676, + "learning_rate": 3.0356526807369624e-05, + "loss": 0.092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04120983928442001, + "step": 5005, + "valid_targets_mean": 3900.0, + "valid_targets_min": 3681 + }, + { + "epoch": 2.7618522601984563, + "grad_norm": 0.39352416278520286, + "learning_rate": 3.0332999462646414e-05, + "loss": 0.1064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050314005464315414, + "step": 5010, + "valid_targets_mean": 2927.0, + "valid_targets_min": 836 + }, + { + "epoch": 2.7646085997794927, + "grad_norm": 0.32843320489684275, + "learning_rate": 3.030945259586385e-05, + "loss": 0.0834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04262343794107437, + "step": 5015, + "valid_targets_mean": 3101.9, + "valid_targets_min": 858 + }, + { + "epoch": 2.767364939360529, + "grad_norm": 0.5602493500282144, + "learning_rate": 3.0285886251508842e-05, + "loss": 0.1064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05516090989112854, + "step": 5020, + "valid_targets_mean": 1645.6, + "valid_targets_min": 512 + }, + { + "epoch": 2.7701212789415655, + "grad_norm": 0.4368539375538447, + "learning_rate": 3.026230047410511e-05, + "loss": 0.1101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044894177466630936, + "step": 5025, + "valid_targets_mean": 2134.5, + "valid_targets_min": 586 + }, + { + "epoch": 2.772877618522602, + "grad_norm": 0.3977141369901912, + "learning_rate": 3.0238695308213086e-05, + "loss": 0.1131, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05769795924425125, + "step": 5030, + "valid_targets_mean": 3461.8, + "valid_targets_min": 2907 + }, + { + "epoch": 2.775633958103638, + "grad_norm": 0.4145007823964718, + "learning_rate": 3.0215070798429837e-05, + "loss": 0.1158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06981924921274185, + "step": 5035, + "valid_targets_mean": 3035.8, + "valid_targets_min": 1335 + }, + { + "epoch": 2.7783902976846746, + "grad_norm": 0.37477133260411005, + "learning_rate": 3.0191426989388985e-05, + "loss": 0.139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05559319257736206, + "step": 5040, + "valid_targets_mean": 3608.2, + "valid_targets_min": 2411 + }, + { + "epoch": 2.781146637265711, + "grad_norm": 0.3698473703132716, + "learning_rate": 3.0167763925760592e-05, + "loss": 0.0943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05161293223500252, + "step": 5045, + "valid_targets_mean": 3239.6, + "valid_targets_min": 665 + }, + { + "epoch": 2.7839029768467474, + "grad_norm": 0.6451322618096782, + "learning_rate": 3.0144081652251107e-05, + "loss": 0.116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07486656308174133, + "step": 5050, + "valid_targets_mean": 1610.0, + "valid_targets_min": 698 + }, + { + "epoch": 2.7866593164277838, + "grad_norm": 0.648461041123343, + "learning_rate": 3.012038021360328e-05, + "loss": 0.1363, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07183019816875458, + "step": 5055, + "valid_targets_mean": 1241.5, + "valid_targets_min": 173 + }, + { + "epoch": 2.78941565600882, + "grad_norm": 0.3509580947065518, + "learning_rate": 3.0096659654596064e-05, + "loss": 0.115, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05083206295967102, + "step": 5060, + "valid_targets_mean": 3774.6, + "valid_targets_min": 2819 + }, + { + "epoch": 2.7921719955898565, + "grad_norm": 0.45344316111335103, + "learning_rate": 3.0072920020044533e-05, + "loss": 0.1533, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14386340975761414, + "step": 5065, + "valid_targets_mean": 3714.9, + "valid_targets_min": 3094 + }, + { + "epoch": 2.794928335170893, + "grad_norm": 0.38840894813200977, + "learning_rate": 3.00491613547998e-05, + "loss": 0.2536, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04627642780542374, + "step": 5070, + "valid_targets_mean": 3465.8, + "valid_targets_min": 1924 + }, + { + "epoch": 2.7976846747519293, + "grad_norm": 0.49517194718843394, + "learning_rate": 3.002538370374894e-05, + "loss": 0.1196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06370240449905396, + "step": 5075, + "valid_targets_mean": 1947.0, + "valid_targets_min": 815 + }, + { + "epoch": 2.8004410143329657, + "grad_norm": 0.41476981506436666, + "learning_rate": 3.000158711181489e-05, + "loss": 0.1123, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07295282930135727, + "step": 5080, + "valid_targets_mean": 2904.6, + "valid_targets_min": 959 + }, + { + "epoch": 2.803197353914002, + "grad_norm": 0.33328045348509194, + "learning_rate": 2.997777162395638e-05, + "loss": 0.1984, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04342009872198105, + "step": 5085, + "valid_targets_mean": 4504.0, + "valid_targets_min": 2253 + }, + { + "epoch": 2.8059536934950384, + "grad_norm": 0.35479127565938806, + "learning_rate": 2.9953937285167827e-05, + "loss": 0.0956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07813682407140732, + "step": 5090, + "valid_targets_mean": 4584.2, + "valid_targets_min": 3860 + }, + { + "epoch": 2.808710033076075, + "grad_norm": 0.35973396064063196, + "learning_rate": 2.993008414047928e-05, + "loss": 0.1043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05732972174882889, + "step": 5095, + "valid_targets_mean": 3729.4, + "valid_targets_min": 510 + }, + { + "epoch": 2.811466372657111, + "grad_norm": 0.30526148170688233, + "learning_rate": 2.9906212234956302e-05, + "loss": 0.1062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0383242703974247, + "step": 5100, + "valid_targets_mean": 4107.1, + "valid_targets_min": 3247 + }, + { + "epoch": 2.8142227122381476, + "grad_norm": 0.31311011239551617, + "learning_rate": 2.9882321613699905e-05, + "loss": 0.0951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047751352190971375, + "step": 5105, + "valid_targets_mean": 4838.2, + "valid_targets_min": 3351 + }, + { + "epoch": 2.8169790518191844, + "grad_norm": 0.3663047510812127, + "learning_rate": 2.9858412321846467e-05, + "loss": 0.0949, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05399296432733536, + "step": 5110, + "valid_targets_mean": 3208.6, + "valid_targets_min": 1621 + }, + { + "epoch": 2.8197353914002203, + "grad_norm": 0.4331457693203822, + "learning_rate": 2.983448440456764e-05, + "loss": 0.0996, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05719532072544098, + "step": 5115, + "valid_targets_mean": 2683.0, + "valid_targets_min": 2126 + }, + { + "epoch": 2.822491730981257, + "grad_norm": 0.39629106488743193, + "learning_rate": 2.9810537907070256e-05, + "loss": 0.0851, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035505153238773346, + "step": 5120, + "valid_targets_mean": 2194.0, + "valid_targets_min": 594 + }, + { + "epoch": 2.825248070562293, + "grad_norm": 0.5187012363545473, + "learning_rate": 2.9786572874596254e-05, + "loss": 0.1525, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07520056515932083, + "step": 5125, + "valid_targets_mean": 2108.0, + "valid_targets_min": 703 + }, + { + "epoch": 2.82800441014333, + "grad_norm": 0.3531803719335357, + "learning_rate": 2.976258935242259e-05, + "loss": 0.0767, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03603310510516167, + "step": 5130, + "valid_targets_mean": 2886.8, + "valid_targets_min": 469 + }, + { + "epoch": 2.830760749724366, + "grad_norm": 0.41473076932503367, + "learning_rate": 2.9738587385861164e-05, + "loss": 0.0927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04555986821651459, + "step": 5135, + "valid_targets_mean": 1476.8, + "valid_targets_min": 613 + }, + { + "epoch": 2.8335170893054027, + "grad_norm": 0.32498394290812244, + "learning_rate": 2.9714567020258712e-05, + "loss": 0.0843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04263223335146904, + "step": 5140, + "valid_targets_mean": 4232.9, + "valid_targets_min": 3120 + }, + { + "epoch": 2.8362734288864386, + "grad_norm": 0.40763953461765295, + "learning_rate": 2.9690528300996725e-05, + "loss": 0.1042, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10327582061290741, + "step": 5145, + "valid_targets_mean": 2615.2, + "valid_targets_min": 760 + }, + { + "epoch": 2.8390297684674755, + "grad_norm": 0.3084220395763072, + "learning_rate": 2.9666471273491382e-05, + "loss": 0.126, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046590656042099, + "step": 5150, + "valid_targets_mean": 3183.9, + "valid_targets_min": 724 + }, + { + "epoch": 2.8417861080485114, + "grad_norm": 0.4991532529087356, + "learning_rate": 2.9642395983193452e-05, + "loss": 0.1294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052444249391555786, + "step": 5155, + "valid_targets_mean": 1333.8, + "valid_targets_min": 685 + }, + { + "epoch": 2.844542447629548, + "grad_norm": 0.6209231754823537, + "learning_rate": 2.961830247558819e-05, + "loss": 0.1127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08437599986791611, + "step": 5160, + "valid_targets_mean": 1587.6, + "valid_targets_min": 660 + }, + { + "epoch": 2.847298787210584, + "grad_norm": 0.4643400060681581, + "learning_rate": 2.9594190796195295e-05, + "loss": 0.1259, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05957096070051193, + "step": 5165, + "valid_targets_mean": 2014.9, + "valid_targets_min": 827 + }, + { + "epoch": 2.850055126791621, + "grad_norm": 0.3672092547777625, + "learning_rate": 2.957006099056878e-05, + "loss": 0.1401, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05223450809717178, + "step": 5170, + "valid_targets_mean": 3783.5, + "valid_targets_min": 2431 + }, + { + "epoch": 2.852811466372657, + "grad_norm": 0.38395979100196775, + "learning_rate": 2.9545913104296898e-05, + "loss": 0.1053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04528707638382912, + "step": 5175, + "valid_targets_mean": 3359.5, + "valid_targets_min": 1344 + }, + { + "epoch": 2.8555678059536937, + "grad_norm": 0.4012102021941841, + "learning_rate": 2.9521747183002086e-05, + "loss": 0.123, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039934054017066956, + "step": 5180, + "valid_targets_mean": 2813.1, + "valid_targets_min": 751 + }, + { + "epoch": 2.8583241455347297, + "grad_norm": 0.407528516968689, + "learning_rate": 2.9497563272340834e-05, + "loss": 0.1081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048174865543842316, + "step": 5185, + "valid_targets_mean": 1752.9, + "valid_targets_min": 482 + }, + { + "epoch": 2.8610804851157665, + "grad_norm": 0.4183370523721934, + "learning_rate": 2.947336141800363e-05, + "loss": 0.0958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0415959395468235, + "step": 5190, + "valid_targets_mean": 3314.1, + "valid_targets_min": 805 + }, + { + "epoch": 2.8638368246968025, + "grad_norm": 0.2553331635874611, + "learning_rate": 2.9449141665714846e-05, + "loss": 0.0956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03737759590148926, + "step": 5195, + "valid_targets_mean": 5540.6, + "valid_targets_min": 5089 + }, + { + "epoch": 2.8665931642778393, + "grad_norm": 0.40080910367296607, + "learning_rate": 2.9424904061232687e-05, + "loss": 0.0959, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059038445353507996, + "step": 5200, + "valid_targets_mean": 3600.2, + "valid_targets_min": 879 + }, + { + "epoch": 2.869349503858875, + "grad_norm": 0.33779405445714833, + "learning_rate": 2.9400648650349094e-05, + "loss": 0.0907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0440572090446949, + "step": 5205, + "valid_targets_mean": 3559.1, + "valid_targets_min": 934 + }, + { + "epoch": 2.872105843439912, + "grad_norm": 0.42172871408078727, + "learning_rate": 2.9376375478889613e-05, + "loss": 0.1075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08035558462142944, + "step": 5210, + "valid_targets_mean": 2765.2, + "valid_targets_min": 804 + }, + { + "epoch": 2.874862183020948, + "grad_norm": 0.41484294086416096, + "learning_rate": 2.9352084592713378e-05, + "loss": 0.1019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05787445977330208, + "step": 5215, + "valid_targets_mean": 3218.0, + "valid_targets_min": 1803 + }, + { + "epoch": 2.877618522601985, + "grad_norm": 0.3873050599704316, + "learning_rate": 2.9327776037712978e-05, + "loss": 0.1101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06193896383047104, + "step": 5220, + "valid_targets_mean": 3538.5, + "valid_targets_min": 2193 + }, + { + "epoch": 2.8803748621830207, + "grad_norm": 0.30413430723900553, + "learning_rate": 2.9303449859814385e-05, + "loss": 0.0912, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04254645109176636, + "step": 5225, + "valid_targets_mean": 3896.2, + "valid_targets_min": 2937 + }, + { + "epoch": 2.8831312017640576, + "grad_norm": 0.4089572111756634, + "learning_rate": 2.927910610497687e-05, + "loss": 0.1003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05655237287282944, + "step": 5230, + "valid_targets_mean": 2332.9, + "valid_targets_min": 880 + }, + { + "epoch": 2.8858875413450935, + "grad_norm": 0.33375196267739643, + "learning_rate": 2.925474481919291e-05, + "loss": 0.1049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04566332697868347, + "step": 5235, + "valid_targets_mean": 3415.6, + "valid_targets_min": 2300 + }, + { + "epoch": 2.8886438809261303, + "grad_norm": 0.34933622872969866, + "learning_rate": 2.9230366048488097e-05, + "loss": 0.0824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04393007606267929, + "step": 5240, + "valid_targets_mean": 3456.6, + "valid_targets_min": 1028 + }, + { + "epoch": 2.8914002205071663, + "grad_norm": 0.33304655007448297, + "learning_rate": 2.9205969838921065e-05, + "loss": 0.1083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04717215895652771, + "step": 5245, + "valid_targets_mean": 2957.1, + "valid_targets_min": 528 + }, + { + "epoch": 2.894156560088203, + "grad_norm": 0.4023147882806599, + "learning_rate": 2.9181556236583385e-05, + "loss": 0.1, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043440282344818115, + "step": 5250, + "valid_targets_mean": 3214.2, + "valid_targets_min": 997 + }, + { + "epoch": 2.896912899669239, + "grad_norm": 0.4529558989658659, + "learning_rate": 2.9157125287599508e-05, + "loss": 0.1184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03613840788602829, + "step": 5255, + "valid_targets_mean": 1525.6, + "valid_targets_min": 594 + }, + { + "epoch": 2.899669239250276, + "grad_norm": 0.4038712949460262, + "learning_rate": 2.9132677038126637e-05, + "loss": 0.1011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03536819666624069, + "step": 5260, + "valid_targets_mean": 1607.8, + "valid_targets_min": 662 + }, + { + "epoch": 2.902425578831312, + "grad_norm": 0.5667316167693416, + "learning_rate": 2.9108211534354672e-05, + "loss": 0.1189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052270498126745224, + "step": 5265, + "valid_targets_mean": 1200.2, + "valid_targets_min": 849 + }, + { + "epoch": 2.9051819184123486, + "grad_norm": 0.4874724217988893, + "learning_rate": 2.9083728822506103e-05, + "loss": 0.1098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07671140134334564, + "step": 5270, + "valid_targets_mean": 1715.2, + "valid_targets_min": 892 + }, + { + "epoch": 2.9079382579933846, + "grad_norm": 0.4854039505754826, + "learning_rate": 2.9059228948835934e-05, + "loss": 0.1043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05655071139335632, + "step": 5275, + "valid_targets_mean": 2259.9, + "valid_targets_min": 735 + }, + { + "epoch": 2.9106945975744214, + "grad_norm": 0.3212040807374135, + "learning_rate": 2.9034711959631607e-05, + "loss": 0.0956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03667246550321579, + "step": 5280, + "valid_targets_mean": 3359.8, + "valid_targets_min": 2814 + }, + { + "epoch": 2.9134509371554573, + "grad_norm": 0.3508251741274065, + "learning_rate": 2.901017790121288e-05, + "loss": 0.0947, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04256372153759003, + "step": 5285, + "valid_targets_mean": 3794.9, + "valid_targets_min": 2873 + }, + { + "epoch": 2.916207276736494, + "grad_norm": 0.49032837443377814, + "learning_rate": 2.898562681993177e-05, + "loss": 0.1243, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055952176451683044, + "step": 5290, + "valid_targets_mean": 1490.1, + "valid_targets_min": 443 + }, + { + "epoch": 2.91896361631753, + "grad_norm": 0.35429788741830825, + "learning_rate": 2.8961058762172448e-05, + "loss": 0.0868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037987325340509415, + "step": 5295, + "valid_targets_mean": 2663.0, + "valid_targets_min": 678 + }, + { + "epoch": 2.921719955898567, + "grad_norm": 0.41811462312373165, + "learning_rate": 2.8936473774351167e-05, + "loss": 0.0962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06996766477823257, + "step": 5300, + "valid_targets_mean": 3080.0, + "valid_targets_min": 795 + }, + { + "epoch": 2.924476295479603, + "grad_norm": 0.35190656538424026, + "learning_rate": 2.891187190291617e-05, + "loss": 0.0959, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048834431916475296, + "step": 5305, + "valid_targets_mean": 4029.0, + "valid_targets_min": 2282 + }, + { + "epoch": 2.9272326350606397, + "grad_norm": 0.3423947169020511, + "learning_rate": 2.888725319434758e-05, + "loss": 0.0832, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034431472420692444, + "step": 5310, + "valid_targets_mean": 3571.1, + "valid_targets_min": 856 + }, + { + "epoch": 2.9299889746416756, + "grad_norm": 0.36953973228180925, + "learning_rate": 2.886261769515734e-05, + "loss": 0.0916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061870966106653214, + "step": 5315, + "valid_targets_mean": 3969.9, + "valid_targets_min": 3010 + }, + { + "epoch": 2.9327453142227125, + "grad_norm": 0.3182169555916361, + "learning_rate": 2.883796545188912e-05, + "loss": 0.0989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059645816683769226, + "step": 5320, + "valid_targets_mean": 3696.4, + "valid_targets_min": 2055 + }, + { + "epoch": 2.9355016538037484, + "grad_norm": 0.38327361537673466, + "learning_rate": 2.8813296511118234e-05, + "loss": 0.1205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05941852182149887, + "step": 5325, + "valid_targets_mean": 3658.6, + "valid_targets_min": 2646 + }, + { + "epoch": 2.938257993384785, + "grad_norm": 0.45780204129304, + "learning_rate": 2.8788610919451504e-05, + "loss": 0.1134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11975133419036865, + "step": 5330, + "valid_targets_mean": 3305.0, + "valid_targets_min": 850 + }, + { + "epoch": 2.941014332965821, + "grad_norm": 0.4378055400695369, + "learning_rate": 2.8763908723527256e-05, + "loss": 0.0831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045288778841495514, + "step": 5335, + "valid_targets_mean": 1481.0, + "valid_targets_min": 811 + }, + { + "epoch": 2.943770672546858, + "grad_norm": 0.3794958177976835, + "learning_rate": 2.8739189970015152e-05, + "loss": 0.1208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04891357570886612, + "step": 5340, + "valid_targets_mean": 3004.9, + "valid_targets_min": 719 + }, + { + "epoch": 2.946527012127894, + "grad_norm": 0.37743104179311454, + "learning_rate": 2.8714454705616157e-05, + "loss": 0.0952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05609768256545067, + "step": 5345, + "valid_targets_mean": 3818.9, + "valid_targets_min": 3374 + }, + { + "epoch": 2.9492833517089307, + "grad_norm": 0.3513156496377298, + "learning_rate": 2.868970297706242e-05, + "loss": 0.1018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04401461035013199, + "step": 5350, + "valid_targets_mean": 3588.9, + "valid_targets_min": 739 + }, + { + "epoch": 2.9520396912899667, + "grad_norm": 0.5440407094274364, + "learning_rate": 2.8664934831117205e-05, + "loss": 0.1134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10201232135295868, + "step": 5355, + "valid_targets_mean": 2562.2, + "valid_targets_min": 837 + }, + { + "epoch": 2.9547960308710035, + "grad_norm": 0.3902171799990619, + "learning_rate": 2.864015031457478e-05, + "loss": 0.1349, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08391343057155609, + "step": 5360, + "valid_targets_mean": 2719.0, + "valid_targets_min": 559 + }, + { + "epoch": 2.9575523704520394, + "grad_norm": 0.43103185077149714, + "learning_rate": 2.8615349474260345e-05, + "loss": 0.1046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040561895817518234, + "step": 5365, + "valid_targets_mean": 2451.2, + "valid_targets_min": 659 + }, + { + "epoch": 2.9603087100330763, + "grad_norm": 0.46689272911503377, + "learning_rate": 2.8590532357029955e-05, + "loss": 0.099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05483695864677429, + "step": 5370, + "valid_targets_mean": 1652.0, + "valid_targets_min": 713 + }, + { + "epoch": 2.963065049614112, + "grad_norm": 0.38076146915136455, + "learning_rate": 2.8565699009770393e-05, + "loss": 0.0952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05026460438966751, + "step": 5375, + "valid_targets_mean": 2261.4, + "valid_targets_min": 539 + }, + { + "epoch": 2.965821389195149, + "grad_norm": 0.3788520764689372, + "learning_rate": 2.8540849479399134e-05, + "loss": 0.0949, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0522831529378891, + "step": 5380, + "valid_targets_mean": 3042.4, + "valid_targets_min": 1014 + }, + { + "epoch": 2.968577728776185, + "grad_norm": 0.3791150587502143, + "learning_rate": 2.8515983812864197e-05, + "loss": 0.0852, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04028245806694031, + "step": 5385, + "valid_targets_mean": 1789.0, + "valid_targets_min": 631 + }, + { + "epoch": 2.971334068357222, + "grad_norm": 0.35207362000832465, + "learning_rate": 2.849110205714411e-05, + "loss": 0.1195, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03954653814435005, + "step": 5390, + "valid_targets_mean": 2367.8, + "valid_targets_min": 790 + }, + { + "epoch": 2.9740904079382577, + "grad_norm": 0.3667392268912815, + "learning_rate": 2.8466204259247786e-05, + "loss": 0.086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03311988711357117, + "step": 5395, + "valid_targets_mean": 1381.9, + "valid_targets_min": 700 + }, + { + "epoch": 2.9768467475192946, + "grad_norm": 0.35776558568368444, + "learning_rate": 2.8441290466214453e-05, + "loss": 0.0964, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03635668009519577, + "step": 5400, + "valid_targets_mean": 2650.1, + "valid_targets_min": 527 + }, + { + "epoch": 2.9796030871003305, + "grad_norm": 0.39315292409803565, + "learning_rate": 2.841636072511355e-05, + "loss": 0.1024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0564381442964077, + "step": 5405, + "valid_targets_mean": 3526.8, + "valid_targets_min": 2880 + }, + { + "epoch": 2.9823594266813673, + "grad_norm": 0.2990859660677176, + "learning_rate": 2.839141508304466e-05, + "loss": 0.0933, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040524035692214966, + "step": 5410, + "valid_targets_mean": 3908.1, + "valid_targets_min": 513 + }, + { + "epoch": 2.9851157662624033, + "grad_norm": 0.2817913429380637, + "learning_rate": 2.8366453587137393e-05, + "loss": 0.0846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03836214542388916, + "step": 5415, + "valid_targets_mean": 3484.6, + "valid_targets_min": 1025 + }, + { + "epoch": 2.98787210584344, + "grad_norm": 0.2760600051991515, + "learning_rate": 2.8341476284551324e-05, + "loss": 0.09, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04450797289609909, + "step": 5420, + "valid_targets_mean": 3447.2, + "valid_targets_min": 601 + }, + { + "epoch": 2.990628445424476, + "grad_norm": 0.48667136142125866, + "learning_rate": 2.8316483222475883e-05, + "loss": 0.1659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16062188148498535, + "step": 5425, + "valid_targets_mean": 2789.1, + "valid_targets_min": 1898 + }, + { + "epoch": 2.993384785005513, + "grad_norm": 0.35165622001090824, + "learning_rate": 2.829147444813028e-05, + "loss": 0.1154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045620642602443695, + "step": 5430, + "valid_targets_mean": 2930.8, + "valid_targets_min": 1228 + }, + { + "epoch": 2.996141124586549, + "grad_norm": 0.3940308674317559, + "learning_rate": 2.82664500087634e-05, + "loss": 0.1016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052383288741111755, + "step": 5435, + "valid_targets_mean": 2870.4, + "valid_targets_min": 713 + }, + { + "epoch": 2.9988974641675856, + "grad_norm": 0.3674963390835259, + "learning_rate": 2.8241409951653732e-05, + "loss": 0.0879, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04892679676413536, + "step": 5440, + "valid_targets_mean": 2235.4, + "valid_targets_min": 1193 + }, + { + "epoch": 3.001653803748622, + "grad_norm": 0.5011492233359499, + "learning_rate": 2.8216354324109284e-05, + "loss": 0.1983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10298144072294235, + "step": 5445, + "valid_targets_mean": 5409.0, + "valid_targets_min": 3500 + }, + { + "epoch": 3.0044101433296584, + "grad_norm": 0.3547841683915858, + "learning_rate": 2.819128317346745e-05, + "loss": 0.2096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10098449140787125, + "step": 5450, + "valid_targets_mean": 7623.9, + "valid_targets_min": 6322 + }, + { + "epoch": 3.0071664829106948, + "grad_norm": 0.358543508375578, + "learning_rate": 2.8166196547094986e-05, + "loss": 0.2004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11547142267227173, + "step": 5455, + "valid_targets_mean": 7311.9, + "valid_targets_min": 5722 + }, + { + "epoch": 3.009922822491731, + "grad_norm": 0.34729115326131915, + "learning_rate": 2.8141094492387853e-05, + "loss": 0.1943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08726060390472412, + "step": 5460, + "valid_targets_mean": 7635.8, + "valid_targets_min": 5634 + }, + { + "epoch": 3.0126791620727675, + "grad_norm": 0.3731758669687285, + "learning_rate": 2.811597705677118e-05, + "loss": 0.1954, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10114558041095734, + "step": 5465, + "valid_targets_mean": 7440.2, + "valid_targets_min": 5614 + }, + { + "epoch": 3.015435501653804, + "grad_norm": 0.3827505449058588, + "learning_rate": 2.8090844287699165e-05, + "loss": 0.1953, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11461259424686432, + "step": 5470, + "valid_targets_mean": 7015.1, + "valid_targets_min": 5539 + }, + { + "epoch": 3.0181918412348403, + "grad_norm": 0.388937271659995, + "learning_rate": 2.806569623265495e-05, + "loss": 0.2106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11077465116977692, + "step": 5475, + "valid_targets_mean": 7799.6, + "valid_targets_min": 5373 + }, + { + "epoch": 3.0209481808158767, + "grad_norm": 0.3650052419451343, + "learning_rate": 2.8040532939150577e-05, + "loss": 0.1978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10374701023101807, + "step": 5480, + "valid_targets_mean": 6996.5, + "valid_targets_min": 5640 + }, + { + "epoch": 3.023704520396913, + "grad_norm": 0.3464200298644989, + "learning_rate": 2.8015354454726863e-05, + "loss": 0.1925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09430479258298874, + "step": 5485, + "valid_targets_mean": 7040.8, + "valid_targets_min": 5071 + }, + { + "epoch": 3.0264608599779494, + "grad_norm": 0.37604560500076617, + "learning_rate": 2.7990160826953336e-05, + "loss": 0.1828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09246242046356201, + "step": 5490, + "valid_targets_mean": 6681.4, + "valid_targets_min": 5437 + }, + { + "epoch": 3.029217199558986, + "grad_norm": 0.34834850520861094, + "learning_rate": 2.796495210342813e-05, + "loss": 0.1929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09243901073932648, + "step": 5495, + "valid_targets_mean": 7258.4, + "valid_targets_min": 5481 + }, + { + "epoch": 3.031973539140022, + "grad_norm": 0.35379365786524, + "learning_rate": 2.7939728331777892e-05, + "loss": 0.186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09171052277088165, + "step": 5500, + "valid_targets_mean": 6993.5, + "valid_targets_min": 5666 + }, + { + "epoch": 3.0347298787210586, + "grad_norm": 0.3662381749119616, + "learning_rate": 2.7914489559657726e-05, + "loss": 0.1955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10121448338031769, + "step": 5505, + "valid_targets_mean": 6706.2, + "valid_targets_min": 5518 + }, + { + "epoch": 3.037486218302095, + "grad_norm": 0.35365025714971077, + "learning_rate": 2.788923583475104e-05, + "loss": 0.1808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08782950043678284, + "step": 5510, + "valid_targets_mean": 6317.2, + "valid_targets_min": 5034 + }, + { + "epoch": 3.0402425578831314, + "grad_norm": 0.35857050025758663, + "learning_rate": 2.786396720476951e-05, + "loss": 0.1865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0956457108259201, + "step": 5515, + "valid_targets_mean": 6752.5, + "valid_targets_min": 5586 + }, + { + "epoch": 3.0429988974641677, + "grad_norm": 0.3785940187599257, + "learning_rate": 2.7838683717452988e-05, + "loss": 0.1927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1081501692533493, + "step": 5520, + "valid_targets_mean": 7536.1, + "valid_targets_min": 5824 + }, + { + "epoch": 3.045755237045204, + "grad_norm": 0.3515804498791697, + "learning_rate": 2.7813385420569364e-05, + "loss": 0.1838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0893816128373146, + "step": 5525, + "valid_targets_mean": 7168.0, + "valid_targets_min": 4761 + }, + { + "epoch": 3.0485115766262405, + "grad_norm": 0.3802559342325645, + "learning_rate": 2.778807236191452e-05, + "loss": 0.1898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0964086502790451, + "step": 5530, + "valid_targets_mean": 6651.6, + "valid_targets_min": 5318 + }, + { + "epoch": 3.051267916207277, + "grad_norm": 0.3817988996444387, + "learning_rate": 2.7762744589312238e-05, + "loss": 0.2051, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10059042274951935, + "step": 5535, + "valid_targets_mean": 6741.1, + "valid_targets_min": 5735 + }, + { + "epoch": 3.0540242557883133, + "grad_norm": 0.7058198957980306, + "learning_rate": 2.7737402150614098e-05, + "loss": 0.2005, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07273004949092865, + "step": 5540, + "valid_targets_mean": 1413.6, + "valid_targets_min": 287 + }, + { + "epoch": 3.0567805953693497, + "grad_norm": 0.43051224485438316, + "learning_rate": 2.7712045093699372e-05, + "loss": 0.1694, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09896133840084076, + "step": 5545, + "valid_targets_mean": 6754.0, + "valid_targets_min": 5024 + }, + { + "epoch": 3.059536934950386, + "grad_norm": 0.3833786175618854, + "learning_rate": 2.768667346647496e-05, + "loss": 0.1829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09074155241250992, + "step": 5550, + "valid_targets_mean": 6746.0, + "valid_targets_min": 3856 + }, + { + "epoch": 3.0622932745314224, + "grad_norm": 0.39945292587828646, + "learning_rate": 2.766128731687529e-05, + "loss": 0.1875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09703618288040161, + "step": 5555, + "valid_targets_mean": 7015.4, + "valid_targets_min": 6101 + }, + { + "epoch": 3.065049614112459, + "grad_norm": 0.3772027369122265, + "learning_rate": 2.763588669286223e-05, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1005934551358223, + "step": 5560, + "valid_targets_mean": 6858.5, + "valid_targets_min": 5844 + }, + { + "epoch": 3.067805953693495, + "grad_norm": 0.3449682284743238, + "learning_rate": 2.761047164242499e-05, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08679661154747009, + "step": 5565, + "valid_targets_mean": 7529.2, + "valid_targets_min": 6113 + }, + { + "epoch": 3.0705622932745316, + "grad_norm": 0.329224016960588, + "learning_rate": 2.758504221358003e-05, + "loss": 0.1776, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0853276252746582, + "step": 5570, + "valid_targets_mean": 8891.4, + "valid_targets_min": 4760 + }, + { + "epoch": 3.073318632855568, + "grad_norm": 0.30006442619540485, + "learning_rate": 2.7559598454370995e-05, + "loss": 0.1636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07512503117322922, + "step": 5575, + "valid_targets_mean": 7368.4, + "valid_targets_min": 5090 + }, + { + "epoch": 3.0760749724366043, + "grad_norm": 0.3195805441643535, + "learning_rate": 2.753414041286858e-05, + "loss": 0.1687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08285166323184967, + "step": 5580, + "valid_targets_mean": 8453.1, + "valid_targets_min": 4010 + }, + { + "epoch": 3.0788313120176407, + "grad_norm": 0.3764879705191803, + "learning_rate": 2.7508668137170473e-05, + "loss": 0.1865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09297159314155579, + "step": 5585, + "valid_targets_mean": 6708.9, + "valid_targets_min": 4625 + }, + { + "epoch": 3.081587651598677, + "grad_norm": 0.3899442157365059, + "learning_rate": 2.748318167540126e-05, + "loss": 0.1768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09663286805152893, + "step": 5590, + "valid_targets_mean": 7523.5, + "valid_targets_min": 5429 + }, + { + "epoch": 3.0843439911797135, + "grad_norm": 0.4218914485548602, + "learning_rate": 2.7457681075712326e-05, + "loss": 0.1969, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11097608506679535, + "step": 5595, + "valid_targets_mean": 7059.1, + "valid_targets_min": 4847 + }, + { + "epoch": 3.08710033076075, + "grad_norm": 0.3770904361726537, + "learning_rate": 2.7432166386281768e-05, + "loss": 0.1997, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08979395031929016, + "step": 5600, + "valid_targets_mean": 6242.9, + "valid_targets_min": 4585 + }, + { + "epoch": 3.0898566703417862, + "grad_norm": 0.3578882009007499, + "learning_rate": 2.7406637655314288e-05, + "loss": 0.1995, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09083861857652664, + "step": 5605, + "valid_targets_mean": 6787.5, + "valid_targets_min": 5493 + }, + { + "epoch": 3.0926130099228226, + "grad_norm": 0.3645082289127681, + "learning_rate": 2.7381094931041134e-05, + "loss": 0.1841, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0917522981762886, + "step": 5610, + "valid_targets_mean": 7336.0, + "valid_targets_min": 4968 + }, + { + "epoch": 3.095369349503859, + "grad_norm": 0.3797382738417148, + "learning_rate": 2.7355538261719994e-05, + "loss": 0.1919, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08494966477155685, + "step": 5615, + "valid_targets_mean": 6181.1, + "valid_targets_min": 3964 + }, + { + "epoch": 3.0981256890848954, + "grad_norm": 0.3701185103299714, + "learning_rate": 2.7329967695634885e-05, + "loss": 0.1912, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0927087739109993, + "step": 5620, + "valid_targets_mean": 8367.6, + "valid_targets_min": 6099 + }, + { + "epoch": 3.1008820286659318, + "grad_norm": 0.36939154366329163, + "learning_rate": 2.7304383281096096e-05, + "loss": 0.1898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09627091884613037, + "step": 5625, + "valid_targets_mean": 7629.2, + "valid_targets_min": 4042 + }, + { + "epoch": 3.103638368246968, + "grad_norm": 0.34462132535963946, + "learning_rate": 2.7278785066440058e-05, + "loss": 0.1924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09351938962936401, + "step": 5630, + "valid_targets_mean": 7061.6, + "valid_targets_min": 5185 + }, + { + "epoch": 3.1063947078280045, + "grad_norm": 0.35205127560908345, + "learning_rate": 2.7253173100029297e-05, + "loss": 0.193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09113998711109161, + "step": 5635, + "valid_targets_mean": 6969.5, + "valid_targets_min": 5329 + }, + { + "epoch": 3.109151047409041, + "grad_norm": 0.37159815439765154, + "learning_rate": 2.7227547430252315e-05, + "loss": 0.1932, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0951208621263504, + "step": 5640, + "valid_targets_mean": 6327.2, + "valid_targets_min": 5204 + }, + { + "epoch": 3.1119073869900773, + "grad_norm": 0.4855868361429094, + "learning_rate": 2.7201908105523496e-05, + "loss": 0.1899, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10221322625875473, + "step": 5645, + "valid_targets_mean": 5908.1, + "valid_targets_min": 4666 + }, + { + "epoch": 3.1146637265711137, + "grad_norm": 1.091136990752759, + "learning_rate": 2.717625517428303e-05, + "loss": 0.1622, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035864755511283875, + "step": 5650, + "valid_targets_mean": 207.1, + "valid_targets_min": 133 + }, + { + "epoch": 3.11742006615215, + "grad_norm": 0.42125949996749595, + "learning_rate": 2.7150588684996803e-05, + "loss": 0.1922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0980990082025528, + "step": 5655, + "valid_targets_mean": 5976.6, + "valid_targets_min": 4924 + }, + { + "epoch": 3.1201764057331864, + "grad_norm": 0.3718325108234843, + "learning_rate": 2.712490868615633e-05, + "loss": 0.1844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08188675343990326, + "step": 5660, + "valid_targets_mean": 6010.1, + "valid_targets_min": 4785 + }, + { + "epoch": 3.122932745314223, + "grad_norm": 0.5996309267330834, + "learning_rate": 2.7099215226278635e-05, + "loss": 0.191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09280208498239517, + "step": 5665, + "valid_targets_mean": 6578.8, + "valid_targets_min": 5070 + }, + { + "epoch": 3.125689084895259, + "grad_norm": 0.4080526563284597, + "learning_rate": 2.707350835390619e-05, + "loss": 0.1838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09484107792377472, + "step": 5670, + "valid_targets_mean": 6638.1, + "valid_targets_min": 4837 + }, + { + "epoch": 3.1284454244762956, + "grad_norm": 0.3799151516797803, + "learning_rate": 2.7047788117606794e-05, + "loss": 0.1834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08327175676822662, + "step": 5675, + "valid_targets_mean": 6058.5, + "valid_targets_min": 4914 + }, + { + "epoch": 3.131201764057332, + "grad_norm": 0.38252610522354596, + "learning_rate": 2.7022054565973498e-05, + "loss": 0.1924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0891115814447403, + "step": 5680, + "valid_targets_mean": 6196.4, + "valid_targets_min": 4824 + }, + { + "epoch": 3.1339581036383684, + "grad_norm": 0.3983212569292108, + "learning_rate": 2.6996307747624514e-05, + "loss": 0.1927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09066662192344666, + "step": 5685, + "valid_targets_mean": 5347.9, + "valid_targets_min": 4122 + }, + { + "epoch": 3.1367144432194047, + "grad_norm": 0.439546583864353, + "learning_rate": 2.6970547711203114e-05, + "loss": 0.1578, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09444321691989899, + "step": 5690, + "valid_targets_mean": 6277.8, + "valid_targets_min": 5273 + }, + { + "epoch": 3.139470782800441, + "grad_norm": 0.4027773477031233, + "learning_rate": 2.6944774505377545e-05, + "loss": 0.1841, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08968262374401093, + "step": 5695, + "valid_targets_mean": 6190.9, + "valid_targets_min": 4605 + }, + { + "epoch": 3.1422271223814775, + "grad_norm": 0.3762342002688076, + "learning_rate": 2.6918988178840932e-05, + "loss": 0.1866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09808917343616486, + "step": 5700, + "valid_targets_mean": 6797.0, + "valid_targets_min": 5636 + }, + { + "epoch": 3.144983461962514, + "grad_norm": 0.39117717729706736, + "learning_rate": 2.6893188780311182e-05, + "loss": 0.1875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09867388010025024, + "step": 5705, + "valid_targets_mean": 6405.5, + "valid_targets_min": 4972 + }, + { + "epoch": 3.1477398015435503, + "grad_norm": 0.38339425447690356, + "learning_rate": 2.686737635853093e-05, + "loss": 0.179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0870504155755043, + "step": 5710, + "valid_targets_mean": 6174.4, + "valid_targets_min": 5309 + }, + { + "epoch": 3.1504961411245866, + "grad_norm": 0.4256301571683098, + "learning_rate": 2.6841550962267376e-05, + "loss": 0.1854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09853317588567734, + "step": 5715, + "valid_targets_mean": 6717.0, + "valid_targets_min": 5331 + }, + { + "epoch": 3.153252480705623, + "grad_norm": 0.395538124959118, + "learning_rate": 2.681571264031225e-05, + "loss": 0.1827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0852356031537056, + "step": 5720, + "valid_targets_mean": 5116.1, + "valid_targets_min": 580 + }, + { + "epoch": 3.1560088202866594, + "grad_norm": 0.7117275230064689, + "learning_rate": 2.678986144148171e-05, + "loss": 0.1895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07064900547266006, + "step": 5725, + "valid_targets_mean": 1063.6, + "valid_targets_min": 545 + }, + { + "epoch": 3.158765159867696, + "grad_norm": 0.792836887415968, + "learning_rate": 2.6763997414616226e-05, + "loss": 0.1704, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09825370460748672, + "step": 5730, + "valid_targets_mean": 1494.2, + "valid_targets_min": 493 + }, + { + "epoch": 3.161521499448732, + "grad_norm": 0.8472983152596772, + "learning_rate": 2.673812060858052e-05, + "loss": 0.1668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08118118345737457, + "step": 5735, + "valid_targets_mean": 1162.5, + "valid_targets_min": 495 + }, + { + "epoch": 3.1642778390297686, + "grad_norm": 0.8092904222588784, + "learning_rate": 2.6712231072263453e-05, + "loss": 0.1601, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07587683945894241, + "step": 5740, + "valid_targets_mean": 1161.5, + "valid_targets_min": 513 + }, + { + "epoch": 3.167034178610805, + "grad_norm": 0.9646390117381839, + "learning_rate": 2.6686328854577923e-05, + "loss": 0.1694, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07207523286342621, + "step": 5745, + "valid_targets_mean": 1186.0, + "valid_targets_min": 595 + }, + { + "epoch": 3.1697905181918413, + "grad_norm": 0.7832187966532755, + "learning_rate": 2.6660414004460806e-05, + "loss": 0.1666, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10817532241344452, + "step": 5750, + "valid_targets_mean": 1532.9, + "valid_targets_min": 498 + }, + { + "epoch": 3.1725468577728777, + "grad_norm": 0.6907258463477448, + "learning_rate": 2.663448657087283e-05, + "loss": 0.1605, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08169940114021301, + "step": 5755, + "valid_targets_mean": 1443.2, + "valid_targets_min": 637 + }, + { + "epoch": 3.175303197353914, + "grad_norm": 0.8720413774781023, + "learning_rate": 2.6608546602798512e-05, + "loss": 0.1808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06673717498779297, + "step": 5760, + "valid_targets_mean": 1022.5, + "valid_targets_min": 607 + }, + { + "epoch": 3.1780595369349505, + "grad_norm": 0.8330802534646491, + "learning_rate": 2.6582594149246035e-05, + "loss": 0.1621, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09219291806221008, + "step": 5765, + "valid_targets_mean": 1582.8, + "valid_targets_min": 600 + }, + { + "epoch": 3.180815876515987, + "grad_norm": 0.6644995965899293, + "learning_rate": 2.6556629259247175e-05, + "loss": 0.1672, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07810449600219727, + "step": 5770, + "valid_targets_mean": 1341.9, + "valid_targets_min": 572 + }, + { + "epoch": 3.1835722160970232, + "grad_norm": 0.801942799942342, + "learning_rate": 2.6530651981857217e-05, + "loss": 0.1632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11833561956882477, + "step": 5775, + "valid_targets_mean": 1436.4, + "valid_targets_min": 590 + }, + { + "epoch": 3.1863285556780596, + "grad_norm": 0.7363195386324293, + "learning_rate": 2.6504662366154826e-05, + "loss": 0.1632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08267237991094589, + "step": 5780, + "valid_targets_mean": 1325.0, + "valid_targets_min": 547 + }, + { + "epoch": 3.189084895259096, + "grad_norm": 0.7237687045105251, + "learning_rate": 2.6478660461242e-05, + "loss": 0.1563, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05040174722671509, + "step": 5785, + "valid_targets_mean": 985.8, + "valid_targets_min": 550 + }, + { + "epoch": 3.1918412348401324, + "grad_norm": 0.7022205152528413, + "learning_rate": 2.6452646316243943e-05, + "loss": 0.1452, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06293104588985443, + "step": 5790, + "valid_targets_mean": 1114.9, + "valid_targets_min": 640 + }, + { + "epoch": 3.1945975744211688, + "grad_norm": 0.7850683341513558, + "learning_rate": 2.6426619980308988e-05, + "loss": 0.1537, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08024225383996964, + "step": 5795, + "valid_targets_mean": 1221.4, + "valid_targets_min": 512 + }, + { + "epoch": 3.197353914002205, + "grad_norm": 0.7919971892277234, + "learning_rate": 2.640058150260849e-05, + "loss": 0.16, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08759787678718567, + "step": 5800, + "valid_targets_mean": 1309.2, + "valid_targets_min": 511 + }, + { + "epoch": 3.2001102535832415, + "grad_norm": 0.7536453955052288, + "learning_rate": 2.6374530932336763e-05, + "loss": 0.1668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10328494012355804, + "step": 5805, + "valid_targets_mean": 1517.9, + "valid_targets_min": 502 + }, + { + "epoch": 3.202866593164278, + "grad_norm": 0.754922587699055, + "learning_rate": 2.6348468318710952e-05, + "loss": 0.164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08109268546104431, + "step": 5810, + "valid_targets_mean": 1190.5, + "valid_targets_min": 505 + }, + { + "epoch": 3.2056229327453143, + "grad_norm": 0.7490475093937684, + "learning_rate": 2.6322393710970963e-05, + "loss": 0.1516, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06469984352588654, + "step": 5815, + "valid_targets_mean": 1053.4, + "valid_targets_min": 623 + }, + { + "epoch": 3.2083792723263507, + "grad_norm": 0.742546060012992, + "learning_rate": 2.6296307158379356e-05, + "loss": 0.163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11054123938083649, + "step": 5820, + "valid_targets_mean": 2130.4, + "valid_targets_min": 676 + }, + { + "epoch": 3.211135611907387, + "grad_norm": 0.7918452644540408, + "learning_rate": 2.6270208710221253e-05, + "loss": 0.1707, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11718101799488068, + "step": 5825, + "valid_targets_mean": 1979.2, + "valid_targets_min": 1070 + }, + { + "epoch": 3.2138919514884234, + "grad_norm": 0.7144941073593385, + "learning_rate": 2.6244098415804272e-05, + "loss": 0.1618, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09609626978635788, + "step": 5830, + "valid_targets_mean": 1480.1, + "valid_targets_min": 702 + }, + { + "epoch": 3.21664829106946, + "grad_norm": 1.4665870489025734, + "learning_rate": 2.62179763244584e-05, + "loss": 0.1539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06281956285238266, + "step": 5835, + "valid_targets_mean": 1019.5, + "valid_targets_min": 541 + }, + { + "epoch": 3.219404630650496, + "grad_norm": 0.8955674691424017, + "learning_rate": 2.61918424855359e-05, + "loss": 0.1549, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09279768168926239, + "step": 5840, + "valid_targets_mean": 1312.5, + "valid_targets_min": 737 + }, + { + "epoch": 3.2221609702315326, + "grad_norm": 0.7517947049192855, + "learning_rate": 2.616569694841125e-05, + "loss": 0.1593, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0847194641828537, + "step": 5845, + "valid_targets_mean": 1507.2, + "valid_targets_min": 626 + }, + { + "epoch": 3.224917309812569, + "grad_norm": 0.7360956984828629, + "learning_rate": 2.6139539762481007e-05, + "loss": 0.1624, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08378203213214874, + "step": 5850, + "valid_targets_mean": 1391.4, + "valid_targets_min": 1072 + }, + { + "epoch": 3.2276736493936053, + "grad_norm": 0.7554668382862645, + "learning_rate": 2.6113370977163762e-05, + "loss": 0.1592, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08441130071878433, + "step": 5855, + "valid_targets_mean": 1751.5, + "valid_targets_min": 858 + }, + { + "epoch": 3.2304299889746417, + "grad_norm": 0.7315626571594799, + "learning_rate": 2.6087190641900005e-05, + "loss": 0.1601, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0891822949051857, + "step": 5860, + "valid_targets_mean": 1601.6, + "valid_targets_min": 1041 + }, + { + "epoch": 3.233186328555678, + "grad_norm": 0.8231038558984789, + "learning_rate": 2.6060998806152045e-05, + "loss": 0.1542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0891842395067215, + "step": 5865, + "valid_targets_mean": 1529.4, + "valid_targets_min": 591 + }, + { + "epoch": 3.2359426681367145, + "grad_norm": 1.0310193470026732, + "learning_rate": 2.6034795519403927e-05, + "loss": 0.1532, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061771877110004425, + "step": 5870, + "valid_targets_mean": 977.6, + "valid_targets_min": 455 + }, + { + "epoch": 3.238699007717751, + "grad_norm": 0.7610975810433076, + "learning_rate": 2.600858083116132e-05, + "loss": 0.1636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08365028351545334, + "step": 5875, + "valid_targets_mean": 1353.4, + "valid_targets_min": 883 + }, + { + "epoch": 3.2414553472987873, + "grad_norm": 0.7410550703354929, + "learning_rate": 2.598235479095145e-05, + "loss": 0.1522, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06863072514533997, + "step": 5880, + "valid_targets_mean": 1224.0, + "valid_targets_min": 555 + }, + { + "epoch": 3.2442116868798236, + "grad_norm": 0.7389477964380543, + "learning_rate": 2.5956117448322972e-05, + "loss": 0.1537, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08163061738014221, + "step": 5885, + "valid_targets_mean": 1349.1, + "valid_targets_min": 1041 + }, + { + "epoch": 3.24696802646086, + "grad_norm": 0.755657293646944, + "learning_rate": 2.592986885284591e-05, + "loss": 0.144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07606132328510284, + "step": 5890, + "valid_targets_mean": 1199.8, + "valid_targets_min": 634 + }, + { + "epoch": 3.2497243660418964, + "grad_norm": 0.7274167285319056, + "learning_rate": 2.590360905411154e-05, + "loss": 0.1593, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07285628467798233, + "step": 5895, + "valid_targets_mean": 1329.1, + "valid_targets_min": 671 + }, + { + "epoch": 3.252480705622933, + "grad_norm": 0.7219283165502001, + "learning_rate": 2.5877338101732293e-05, + "loss": 0.1501, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06267837435007095, + "step": 5900, + "valid_targets_mean": 1099.9, + "valid_targets_min": 666 + }, + { + "epoch": 3.255237045203969, + "grad_norm": 0.7786782059656266, + "learning_rate": 2.5851056045341704e-05, + "loss": 0.1536, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08118004351854324, + "step": 5905, + "valid_targets_mean": 1315.5, + "valid_targets_min": 655 + }, + { + "epoch": 3.2579933847850056, + "grad_norm": 0.7901036661025802, + "learning_rate": 2.5824762934594256e-05, + "loss": 0.1503, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08546371757984161, + "step": 5910, + "valid_targets_mean": 1391.0, + "valid_targets_min": 593 + }, + { + "epoch": 3.260749724366042, + "grad_norm": 0.8284137362054649, + "learning_rate": 2.5798458819165335e-05, + "loss": 0.1624, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08809350430965424, + "step": 5915, + "valid_targets_mean": 1355.9, + "valid_targets_min": 716 + }, + { + "epoch": 3.2635060639470783, + "grad_norm": 0.8337027651086962, + "learning_rate": 2.57721437487511e-05, + "loss": 0.1531, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0863991379737854, + "step": 5920, + "valid_targets_mean": 1394.5, + "valid_targets_min": 679 + }, + { + "epoch": 3.2662624035281147, + "grad_norm": 0.6985992477873129, + "learning_rate": 2.5745817773068428e-05, + "loss": 0.1562, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06375578045845032, + "step": 5925, + "valid_targets_mean": 1045.6, + "valid_targets_min": 519 + }, + { + "epoch": 3.269018743109151, + "grad_norm": 0.7646651272406129, + "learning_rate": 2.571948094185479e-05, + "loss": 0.1602, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07174058258533478, + "step": 5930, + "valid_targets_mean": 1214.5, + "valid_targets_min": 810 + }, + { + "epoch": 3.2717750826901875, + "grad_norm": 0.7391296468506471, + "learning_rate": 2.5693133304868164e-05, + "loss": 0.1696, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0897759348154068, + "step": 5935, + "valid_targets_mean": 1373.6, + "valid_targets_min": 475 + }, + { + "epoch": 3.274531422271224, + "grad_norm": 0.9487295946434656, + "learning_rate": 2.5666774911886948e-05, + "loss": 0.1531, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10385408252477646, + "step": 5940, + "valid_targets_mean": 1735.4, + "valid_targets_min": 1195 + }, + { + "epoch": 3.2772877618522602, + "grad_norm": 0.740852652045916, + "learning_rate": 2.5640405812709847e-05, + "loss": 0.1542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05872708931565285, + "step": 5945, + "valid_targets_mean": 1168.1, + "valid_targets_min": 522 + }, + { + "epoch": 3.2800441014332966, + "grad_norm": 0.8271033312223365, + "learning_rate": 2.5614026057155822e-05, + "loss": 0.1584, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10197576880455017, + "step": 5950, + "valid_targets_mean": 1617.0, + "valid_targets_min": 632 + }, + { + "epoch": 3.282800441014333, + "grad_norm": 0.7077108251198051, + "learning_rate": 2.558763569506394e-05, + "loss": 0.1501, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07479213178157806, + "step": 5955, + "valid_targets_mean": 1278.2, + "valid_targets_min": 591 + }, + { + "epoch": 3.2855567805953694, + "grad_norm": 0.7598704496811521, + "learning_rate": 2.556123477629332e-05, + "loss": 0.1496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08798534423112869, + "step": 5960, + "valid_targets_mean": 1560.5, + "valid_targets_min": 955 + }, + { + "epoch": 3.2883131201764058, + "grad_norm": 0.791215530810002, + "learning_rate": 2.5534823350723022e-05, + "loss": 0.146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06130689010024071, + "step": 5965, + "valid_targets_mean": 945.0, + "valid_targets_min": 471 + }, + { + "epoch": 3.291069459757442, + "grad_norm": 0.8145998943809722, + "learning_rate": 2.550840146825195e-05, + "loss": 0.1621, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08687254041433334, + "step": 5970, + "valid_targets_mean": 1331.1, + "valid_targets_min": 689 + }, + { + "epoch": 3.2938257993384785, + "grad_norm": 0.753027273278776, + "learning_rate": 2.548196917879878e-05, + "loss": 0.1564, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08908244967460632, + "step": 5975, + "valid_targets_mean": 1538.2, + "valid_targets_min": 1218 + }, + { + "epoch": 3.296582138919515, + "grad_norm": 0.7164463382195823, + "learning_rate": 2.5455526532301838e-05, + "loss": 0.1525, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0683460384607315, + "step": 5980, + "valid_targets_mean": 1246.2, + "valid_targets_min": 610 + }, + { + "epoch": 3.2993384785005513, + "grad_norm": 0.7708182218186814, + "learning_rate": 2.5429073578719018e-05, + "loss": 0.1514, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07496686279773712, + "step": 5985, + "valid_targets_mean": 1213.4, + "valid_targets_min": 911 + }, + { + "epoch": 3.3020948180815877, + "grad_norm": 0.838791358824481, + "learning_rate": 2.540261036802769e-05, + "loss": 0.1553, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07214908301830292, + "step": 5990, + "valid_targets_mean": 1189.4, + "valid_targets_min": 738 + }, + { + "epoch": 3.304851157662624, + "grad_norm": 0.7105297304688872, + "learning_rate": 2.53761369502246e-05, + "loss": 0.158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09414061903953552, + "step": 5995, + "valid_targets_mean": 1783.5, + "valid_targets_min": 877 + }, + { + "epoch": 3.3076074972436604, + "grad_norm": 0.8213619188326613, + "learning_rate": 2.5349653375325788e-05, + "loss": 0.1526, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07752993702888489, + "step": 6000, + "valid_targets_mean": 1085.1, + "valid_targets_min": 453 + }, + { + "epoch": 3.310363836824697, + "grad_norm": 0.7777825719637373, + "learning_rate": 2.5323159693366472e-05, + "loss": 0.1533, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05202046036720276, + "step": 6005, + "valid_targets_mean": 903.9, + "valid_targets_min": 646 + }, + { + "epoch": 3.313120176405733, + "grad_norm": 0.7503930461724283, + "learning_rate": 2.529665595440096e-05, + "loss": 0.1511, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06798464059829712, + "step": 6010, + "valid_targets_mean": 1135.2, + "valid_targets_min": 686 + }, + { + "epoch": 3.3158765159867696, + "grad_norm": 0.7632551510936336, + "learning_rate": 2.5270142208502586e-05, + "loss": 0.1576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06699888408184052, + "step": 6015, + "valid_targets_mean": 1154.6, + "valid_targets_min": 505 + }, + { + "epoch": 3.318632855567806, + "grad_norm": 0.801609262824881, + "learning_rate": 2.5243618505763562e-05, + "loss": 0.1619, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07657718658447266, + "step": 6020, + "valid_targets_mean": 1196.2, + "valid_targets_min": 539 + }, + { + "epoch": 3.3213891951488423, + "grad_norm": 0.7697926559171048, + "learning_rate": 2.5217084896294926e-05, + "loss": 0.1474, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07430486381053925, + "step": 6025, + "valid_targets_mean": 1239.1, + "valid_targets_min": 513 + }, + { + "epoch": 3.3241455347298787, + "grad_norm": 0.7692663539946475, + "learning_rate": 2.5190541430226432e-05, + "loss": 0.1553, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07357966899871826, + "step": 6030, + "valid_targets_mean": 1123.2, + "valid_targets_min": 559 + }, + { + "epoch": 3.326901874310915, + "grad_norm": 0.78292306650751, + "learning_rate": 2.516398815770645e-05, + "loss": 0.1571, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07210767269134521, + "step": 6035, + "valid_targets_mean": 1243.0, + "valid_targets_min": 727 + }, + { + "epoch": 3.3296582138919515, + "grad_norm": 0.7527434708707634, + "learning_rate": 2.513742512890189e-05, + "loss": 0.1585, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09000750631093979, + "step": 6040, + "valid_targets_mean": 1650.1, + "valid_targets_min": 598 + }, + { + "epoch": 3.332414553472988, + "grad_norm": 0.7987608296922358, + "learning_rate": 2.511085239399807e-05, + "loss": 0.1609, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08733777701854706, + "step": 6045, + "valid_targets_mean": 1339.9, + "valid_targets_min": 571 + }, + { + "epoch": 3.3351708930540243, + "grad_norm": 0.8295624821665083, + "learning_rate": 2.5084270003198672e-05, + "loss": 0.156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09512234479188919, + "step": 6050, + "valid_targets_mean": 1349.6, + "valid_targets_min": 781 + }, + { + "epoch": 3.3379272326350606, + "grad_norm": 0.7525015886815003, + "learning_rate": 2.5057678006725613e-05, + "loss": 0.1586, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07326331734657288, + "step": 6055, + "valid_targets_mean": 1193.4, + "valid_targets_min": 659 + }, + { + "epoch": 3.340683572216097, + "grad_norm": 0.8238463048400307, + "learning_rate": 2.503107645481895e-05, + "loss": 0.1589, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08552856743335724, + "step": 6060, + "valid_targets_mean": 1430.1, + "valid_targets_min": 600 + }, + { + "epoch": 3.3434399117971334, + "grad_norm": 0.7222152955628499, + "learning_rate": 2.5004465397736794e-05, + "loss": 0.1639, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0757959634065628, + "step": 6065, + "valid_targets_mean": 1397.1, + "valid_targets_min": 987 + }, + { + "epoch": 3.34619625137817, + "grad_norm": 0.7403365551416695, + "learning_rate": 2.4977844885755228e-05, + "loss": 0.1607, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08899972587823868, + "step": 6070, + "valid_targets_mean": 1489.4, + "valid_targets_min": 912 + }, + { + "epoch": 3.348952590959206, + "grad_norm": 0.7556423085102235, + "learning_rate": 2.4951214969168177e-05, + "loss": 0.1593, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06914782524108887, + "step": 6075, + "valid_targets_mean": 1080.8, + "valid_targets_min": 465 + }, + { + "epoch": 3.3517089305402425, + "grad_norm": 0.7947196215615459, + "learning_rate": 2.492457569828736e-05, + "loss": 0.1564, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08713622391223907, + "step": 6080, + "valid_targets_mean": 1272.0, + "valid_targets_min": 726 + }, + { + "epoch": 3.354465270121279, + "grad_norm": 0.8081245108137323, + "learning_rate": 2.4897927123442135e-05, + "loss": 0.1518, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08181916922330856, + "step": 6085, + "valid_targets_mean": 1407.4, + "valid_targets_min": 838 + }, + { + "epoch": 3.3572216097023153, + "grad_norm": 0.893548722873977, + "learning_rate": 2.4871269294979473e-05, + "loss": 0.1628, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07865282148122787, + "step": 6090, + "valid_targets_mean": 984.5, + "valid_targets_min": 596 + }, + { + "epoch": 3.3599779492833517, + "grad_norm": 0.790098860007022, + "learning_rate": 2.4844602263263805e-05, + "loss": 0.1515, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061258696019649506, + "step": 6095, + "valid_targets_mean": 1038.1, + "valid_targets_min": 497 + }, + { + "epoch": 3.362734288864388, + "grad_norm": 0.6959492154594862, + "learning_rate": 2.481792607867696e-05, + "loss": 0.1495, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06753015518188477, + "step": 6100, + "valid_targets_mean": 1178.6, + "valid_targets_min": 833 + }, + { + "epoch": 3.3654906284454245, + "grad_norm": 0.7225310027236117, + "learning_rate": 2.479124079161805e-05, + "loss": 0.1529, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07265844941139221, + "step": 6105, + "valid_targets_mean": 1203.9, + "valid_targets_min": 539 + }, + { + "epoch": 3.368246968026461, + "grad_norm": 0.8529450916988595, + "learning_rate": 2.4764546452503394e-05, + "loss": 0.1677, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07796834409236908, + "step": 6110, + "valid_targets_mean": 1061.2, + "valid_targets_min": 546 + }, + { + "epoch": 3.3710033076074972, + "grad_norm": 0.7175824124954218, + "learning_rate": 2.4737843111766407e-05, + "loss": 0.1554, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08355279266834259, + "step": 6115, + "valid_targets_mean": 1195.5, + "valid_targets_min": 488 + }, + { + "epoch": 3.3737596471885336, + "grad_norm": 0.6829589487717951, + "learning_rate": 2.471113081985752e-05, + "loss": 0.156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08508077263832092, + "step": 6120, + "valid_targets_mean": 1609.6, + "valid_targets_min": 769 + }, + { + "epoch": 3.37651598676957, + "grad_norm": 0.6730368478570894, + "learning_rate": 2.4684409627244064e-05, + "loss": 0.1403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05498487502336502, + "step": 6125, + "valid_targets_mean": 1292.5, + "valid_targets_min": 496 + }, + { + "epoch": 3.3792723263506064, + "grad_norm": 0.8046453370147167, + "learning_rate": 2.4657679584410187e-05, + "loss": 0.1454, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07212387025356293, + "step": 6130, + "valid_targets_mean": 1189.5, + "valid_targets_min": 986 + }, + { + "epoch": 3.3820286659316428, + "grad_norm": 0.7295647392573871, + "learning_rate": 2.4630940741856762e-05, + "loss": 0.1549, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05706660449504852, + "step": 6135, + "valid_targets_mean": 1152.4, + "valid_targets_min": 431 + }, + { + "epoch": 3.384785005512679, + "grad_norm": 0.7162755490585064, + "learning_rate": 2.4604193150101287e-05, + "loss": 0.1525, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06993607431650162, + "step": 6140, + "valid_targets_mean": 1334.4, + "valid_targets_min": 585 + }, + { + "epoch": 3.3875413450937155, + "grad_norm": 0.7332666144427297, + "learning_rate": 2.4577436859677795e-05, + "loss": 0.1534, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06883664429187775, + "step": 6145, + "valid_targets_mean": 1118.4, + "valid_targets_min": 704 + }, + { + "epoch": 3.390297684674752, + "grad_norm": 0.680550109871262, + "learning_rate": 2.455067192113674e-05, + "loss": 0.1589, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07858337461948395, + "step": 6150, + "valid_targets_mean": 1383.4, + "valid_targets_min": 471 + }, + { + "epoch": 3.3930540242557883, + "grad_norm": 0.7681284346980127, + "learning_rate": 2.4523898385044924e-05, + "loss": 0.1552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07727697491645813, + "step": 6155, + "valid_targets_mean": 1158.0, + "valid_targets_min": 632 + }, + { + "epoch": 3.3958103638368247, + "grad_norm": 0.8163411240896142, + "learning_rate": 2.449711630198539e-05, + "loss": 0.1691, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1454894095659256, + "step": 6160, + "valid_targets_mean": 2179.1, + "valid_targets_min": 915 + }, + { + "epoch": 3.398566703417861, + "grad_norm": 0.5659264702955422, + "learning_rate": 2.447032572255733e-05, + "loss": 0.267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12513965368270874, + "step": 6165, + "valid_targets_mean": 4942.2, + "valid_targets_min": 455 + }, + { + "epoch": 3.4013230429988974, + "grad_norm": 0.44053618011580853, + "learning_rate": 2.4443526697375987e-05, + "loss": 0.2486, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12482090294361115, + "step": 6170, + "valid_targets_mean": 5150.0, + "valid_targets_min": 1617 + }, + { + "epoch": 3.404079382579934, + "grad_norm": 0.4351829615720785, + "learning_rate": 2.441671927707256e-05, + "loss": 0.2399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08835594356060028, + "step": 6175, + "valid_targets_mean": 4582.6, + "valid_targets_min": 905 + }, + { + "epoch": 3.40683572216097, + "grad_norm": 0.4415519515511239, + "learning_rate": 2.4389903512294115e-05, + "loss": 0.2407, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12533210217952728, + "step": 6180, + "valid_targets_mean": 5638.1, + "valid_targets_min": 2173 + }, + { + "epoch": 3.4095920617420066, + "grad_norm": 0.5543533730075603, + "learning_rate": 2.4363079453703467e-05, + "loss": 0.2503, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10972462594509125, + "step": 6185, + "valid_targets_mean": 3896.4, + "valid_targets_min": 1323 + }, + { + "epoch": 3.412348401323043, + "grad_norm": 0.507729188791909, + "learning_rate": 2.4336247151979128e-05, + "loss": 0.2466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12650559842586517, + "step": 6190, + "valid_targets_mean": 5116.5, + "valid_targets_min": 1307 + }, + { + "epoch": 3.4151047409040793, + "grad_norm": 0.5670545458385622, + "learning_rate": 2.4309406657815146e-05, + "loss": 0.2416, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10593944042921066, + "step": 6195, + "valid_targets_mean": 2957.5, + "valid_targets_min": 1306 + }, + { + "epoch": 3.4178610804851157, + "grad_norm": 0.5103614356087112, + "learning_rate": 2.4282558021921086e-05, + "loss": 0.2435, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10605885833501816, + "step": 6200, + "valid_targets_mean": 4416.8, + "valid_targets_min": 1165 + }, + { + "epoch": 3.420617420066152, + "grad_norm": 0.5307792228138946, + "learning_rate": 2.425570129502187e-05, + "loss": 0.2529, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10670338571071625, + "step": 6205, + "valid_targets_mean": 4066.8, + "valid_targets_min": 981 + }, + { + "epoch": 3.4233737596471885, + "grad_norm": 0.5199931605827699, + "learning_rate": 2.4228836527857705e-05, + "loss": 0.2461, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.131082221865654, + "step": 6210, + "valid_targets_mean": 4610.1, + "valid_targets_min": 2306 + }, + { + "epoch": 3.426130099228225, + "grad_norm": 0.45774349677177545, + "learning_rate": 2.4201963771184008e-05, + "loss": 0.2419, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11225003004074097, + "step": 6215, + "valid_targets_mean": 5402.1, + "valid_targets_min": 1918 + }, + { + "epoch": 3.4288864388092613, + "grad_norm": 0.5131328341021459, + "learning_rate": 2.4175083075771283e-05, + "loss": 0.2391, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11914755403995514, + "step": 6220, + "valid_targets_mean": 4865.1, + "valid_targets_min": 2332 + }, + { + "epoch": 3.4316427783902976, + "grad_norm": 0.4799037716861963, + "learning_rate": 2.4148194492405013e-05, + "loss": 0.248, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12626518309116364, + "step": 6225, + "valid_targets_mean": 4937.5, + "valid_targets_min": 892 + }, + { + "epoch": 3.434399117971334, + "grad_norm": 0.5740445467780841, + "learning_rate": 2.4121298071885613e-05, + "loss": 0.2393, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11141481250524521, + "step": 6230, + "valid_targets_mean": 3301.6, + "valid_targets_min": 1441 + }, + { + "epoch": 3.4371554575523704, + "grad_norm": 0.6039338696728922, + "learning_rate": 2.409439386502828e-05, + "loss": 0.2495, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10693929344415665, + "step": 6235, + "valid_targets_mean": 2775.6, + "valid_targets_min": 524 + }, + { + "epoch": 3.439911797133407, + "grad_norm": 0.5650629644965303, + "learning_rate": 2.4067481922662945e-05, + "loss": 0.2301, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11332054436206818, + "step": 6240, + "valid_targets_mean": 3284.8, + "valid_targets_min": 879 + }, + { + "epoch": 3.442668136714443, + "grad_norm": 0.5900400029918996, + "learning_rate": 2.4040562295634137e-05, + "loss": 0.2391, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0969422310590744, + "step": 6245, + "valid_targets_mean": 3035.5, + "valid_targets_min": 962 + }, + { + "epoch": 3.4454244762954795, + "grad_norm": 0.5838118730511096, + "learning_rate": 2.4013635034800897e-05, + "loss": 0.2489, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12656308710575104, + "step": 6250, + "valid_targets_mean": 4092.2, + "valid_targets_min": 1949 + }, + { + "epoch": 3.448180815876516, + "grad_norm": 0.585636070746801, + "learning_rate": 2.398670019103672e-05, + "loss": 0.2473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10216294974088669, + "step": 6255, + "valid_targets_mean": 3153.2, + "valid_targets_min": 1160 + }, + { + "epoch": 3.4509371554575523, + "grad_norm": 0.5746746563402606, + "learning_rate": 2.395975781522938e-05, + "loss": 0.2453, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1407299041748047, + "step": 6260, + "valid_targets_mean": 4303.6, + "valid_targets_min": 1915 + }, + { + "epoch": 3.4536934950385887, + "grad_norm": 0.560681950637597, + "learning_rate": 2.3932807958280938e-05, + "loss": 0.2295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1079631820321083, + "step": 6265, + "valid_targets_mean": 3943.8, + "valid_targets_min": 1998 + }, + { + "epoch": 3.456449834619625, + "grad_norm": 0.5471776806097317, + "learning_rate": 2.390585067110754e-05, + "loss": 0.2394, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09487134963274002, + "step": 6270, + "valid_targets_mean": 3306.5, + "valid_targets_min": 2369 + }, + { + "epoch": 3.4592061742006615, + "grad_norm": 0.6126534909782733, + "learning_rate": 2.38788860046394e-05, + "loss": 0.2173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12752440571784973, + "step": 6275, + "valid_targets_mean": 3309.4, + "valid_targets_min": 1946 + }, + { + "epoch": 3.461962513781698, + "grad_norm": 0.624614397143209, + "learning_rate": 2.3851914009820645e-05, + "loss": 0.2413, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14843641221523285, + "step": 6280, + "valid_targets_mean": 4116.9, + "valid_targets_min": 2834 + }, + { + "epoch": 3.464718853362734, + "grad_norm": 0.5637514800954267, + "learning_rate": 2.3824934737609283e-05, + "loss": 0.2342, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08752866089344025, + "step": 6285, + "valid_targets_mean": 3024.5, + "valid_targets_min": 930 + }, + { + "epoch": 3.4674751929437706, + "grad_norm": 0.5305645967148676, + "learning_rate": 2.3797948238977044e-05, + "loss": 0.2335, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09303529560565948, + "step": 6290, + "valid_targets_mean": 2898.9, + "valid_targets_min": 470 + }, + { + "epoch": 3.470231532524807, + "grad_norm": 0.5941984461089536, + "learning_rate": 2.3770954564909326e-05, + "loss": 0.239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12768758833408356, + "step": 6295, + "valid_targets_mean": 3422.1, + "valid_targets_min": 1051 + }, + { + "epoch": 3.4729878721058434, + "grad_norm": 0.607894001026613, + "learning_rate": 2.3743953766405067e-05, + "loss": 0.2216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.108496755361557, + "step": 6300, + "valid_targets_mean": 3197.9, + "valid_targets_min": 2275 + }, + { + "epoch": 3.4757442116868797, + "grad_norm": 0.5855734261345544, + "learning_rate": 2.3716945894476692e-05, + "loss": 0.2406, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11739534139633179, + "step": 6305, + "valid_targets_mean": 3330.8, + "valid_targets_min": 1195 + }, + { + "epoch": 3.478500551267916, + "grad_norm": 0.6102917053149225, + "learning_rate": 2.3689931000149962e-05, + "loss": 0.2293, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1125403568148613, + "step": 6310, + "valid_targets_mean": 2895.0, + "valid_targets_min": 584 + }, + { + "epoch": 3.4812568908489525, + "grad_norm": 0.6002847779042462, + "learning_rate": 2.3662909134463917e-05, + "loss": 0.2286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07871828973293304, + "step": 6315, + "valid_targets_mean": 2149.6, + "valid_targets_min": 1405 + }, + { + "epoch": 3.484013230429989, + "grad_norm": 0.6078189364769471, + "learning_rate": 2.363588034847077e-05, + "loss": 0.2264, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12055455893278122, + "step": 6320, + "valid_targets_mean": 3423.0, + "valid_targets_min": 1642 + }, + { + "epoch": 3.4867695700110253, + "grad_norm": 0.5738891126905435, + "learning_rate": 2.3608844693235807e-05, + "loss": 0.2324, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11131959408521652, + "step": 6325, + "valid_targets_mean": 3915.1, + "valid_targets_min": 1306 + }, + { + "epoch": 3.4895259095920617, + "grad_norm": 0.6623793377526902, + "learning_rate": 2.358180221983729e-05, + "loss": 0.2239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10811597108840942, + "step": 6330, + "valid_targets_mean": 3056.5, + "valid_targets_min": 1134 + }, + { + "epoch": 3.492282249173098, + "grad_norm": 0.594160163383452, + "learning_rate": 2.3554752979366367e-05, + "loss": 0.2339, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08721758425235748, + "step": 6335, + "valid_targets_mean": 2610.4, + "valid_targets_min": 1240 + }, + { + "epoch": 3.4950385887541344, + "grad_norm": 0.6206760795442965, + "learning_rate": 2.352769702292696e-05, + "loss": 0.2403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10366570949554443, + "step": 6340, + "valid_targets_mean": 3440.2, + "valid_targets_min": 1482 + }, + { + "epoch": 3.497794928335171, + "grad_norm": 0.5442533453248503, + "learning_rate": 2.3500634401635695e-05, + "loss": 0.2253, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12482604384422302, + "step": 6345, + "valid_targets_mean": 4187.2, + "valid_targets_min": 1568 + }, + { + "epoch": 3.500551267916207, + "grad_norm": 0.5953924130334132, + "learning_rate": 2.3473565166621778e-05, + "loss": 0.2262, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11123813688755035, + "step": 6350, + "valid_targets_mean": 3236.9, + "valid_targets_min": 787 + }, + { + "epoch": 3.5033076074972436, + "grad_norm": 0.5656650287674254, + "learning_rate": 2.3446489369026912e-05, + "loss": 0.2318, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10359810292720795, + "step": 6355, + "valid_targets_mean": 3726.2, + "valid_targets_min": 1947 + }, + { + "epoch": 3.50606394707828, + "grad_norm": 0.587810747473901, + "learning_rate": 2.3419407060005207e-05, + "loss": 0.2366, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11421269178390503, + "step": 6360, + "valid_targets_mean": 3086.5, + "valid_targets_min": 2152 + }, + { + "epoch": 3.5088202866593163, + "grad_norm": 0.6776618331479065, + "learning_rate": 2.339231829072307e-05, + "loss": 0.2259, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14600136876106262, + "step": 6365, + "valid_targets_mean": 3337.1, + "valid_targets_min": 1374 + }, + { + "epoch": 3.5115766262403527, + "grad_norm": 0.5993980288155875, + "learning_rate": 2.33652231123591e-05, + "loss": 0.2333, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11219732463359833, + "step": 6370, + "valid_targets_mean": 3340.2, + "valid_targets_min": 1167 + }, + { + "epoch": 3.514332965821389, + "grad_norm": 0.6218997217450513, + "learning_rate": 2.3338121576104028e-05, + "loss": 0.2284, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12085797637701035, + "step": 6375, + "valid_targets_mean": 3300.8, + "valid_targets_min": 1440 + }, + { + "epoch": 3.5170893054024255, + "grad_norm": 0.5172162406201946, + "learning_rate": 2.3311013733160585e-05, + "loss": 0.2168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1054062768816948, + "step": 6380, + "valid_targets_mean": 4292.4, + "valid_targets_min": 1980 + }, + { + "epoch": 3.519845644983462, + "grad_norm": 0.6075397343264516, + "learning_rate": 2.3283899634743414e-05, + "loss": 0.2242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10251583158969879, + "step": 6385, + "valid_targets_mean": 3302.4, + "valid_targets_min": 1192 + }, + { + "epoch": 3.5226019845644982, + "grad_norm": 0.5382943683418272, + "learning_rate": 2.3256779332078986e-05, + "loss": 0.2392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13341036438941956, + "step": 6390, + "valid_targets_mean": 4459.2, + "valid_targets_min": 2298 + }, + { + "epoch": 3.5253583241455346, + "grad_norm": 0.6073277392034562, + "learning_rate": 2.322965287640548e-05, + "loss": 0.2263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09832696616649628, + "step": 6395, + "valid_targets_mean": 3120.0, + "valid_targets_min": 529 + }, + { + "epoch": 3.528114663726571, + "grad_norm": 0.5822717275745802, + "learning_rate": 2.3202520318972708e-05, + "loss": 0.2393, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12465708702802658, + "step": 6400, + "valid_targets_mean": 3861.0, + "valid_targets_min": 1645 + }, + { + "epoch": 3.5308710033076074, + "grad_norm": 0.5850378373035345, + "learning_rate": 2.317538171104202e-05, + "loss": 0.227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1155102550983429, + "step": 6405, + "valid_targets_mean": 3875.6, + "valid_targets_min": 743 + }, + { + "epoch": 3.5336273428886438, + "grad_norm": 0.5924089353666543, + "learning_rate": 2.314823710388618e-05, + "loss": 0.2329, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12425795197486877, + "step": 6410, + "valid_targets_mean": 3637.5, + "valid_targets_min": 1007 + }, + { + "epoch": 3.53638368246968, + "grad_norm": 0.6077176803524151, + "learning_rate": 2.312108654878929e-05, + "loss": 0.2228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10154540836811066, + "step": 6415, + "valid_targets_mean": 2488.4, + "valid_targets_min": 1246 + }, + { + "epoch": 3.5391400220507165, + "grad_norm": 0.597470246206563, + "learning_rate": 2.3093930097046697e-05, + "loss": 0.2357, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10589481890201569, + "step": 6420, + "valid_targets_mean": 3084.9, + "valid_targets_min": 1213 + }, + { + "epoch": 3.541896361631753, + "grad_norm": 0.510064852197398, + "learning_rate": 2.3066767799964885e-05, + "loss": 0.2191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09461426734924316, + "step": 6425, + "valid_targets_mean": 3938.0, + "valid_targets_min": 2749 + }, + { + "epoch": 3.5446527012127893, + "grad_norm": 0.5583651875591139, + "learning_rate": 2.303959970886138e-05, + "loss": 0.2175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10345014929771423, + "step": 6430, + "valid_targets_mean": 3994.8, + "valid_targets_min": 2062 + }, + { + "epoch": 3.5474090407938257, + "grad_norm": 0.6155314604138545, + "learning_rate": 2.3012425875064657e-05, + "loss": 0.2188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14093247056007385, + "step": 6435, + "valid_targets_mean": 4283.6, + "valid_targets_min": 462 + }, + { + "epoch": 3.550165380374862, + "grad_norm": 0.5792221807458252, + "learning_rate": 2.298524634991403e-05, + "loss": 0.237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11345086246728897, + "step": 6440, + "valid_targets_mean": 3411.4, + "valid_targets_min": 1001 + }, + { + "epoch": 3.5529217199558984, + "grad_norm": 0.5737247056241034, + "learning_rate": 2.2958061184759586e-05, + "loss": 0.2288, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11991208791732788, + "step": 6445, + "valid_targets_mean": 3392.1, + "valid_targets_min": 1622 + }, + { + "epoch": 3.555678059536935, + "grad_norm": 0.5780171713268059, + "learning_rate": 2.2930870430962055e-05, + "loss": 0.2339, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11348418146371841, + "step": 6450, + "valid_targets_mean": 3478.1, + "valid_targets_min": 1709 + }, + { + "epoch": 3.558434399117971, + "grad_norm": 0.5994740318518521, + "learning_rate": 2.2903674139892717e-05, + "loss": 0.2249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10606072843074799, + "step": 6455, + "valid_targets_mean": 3584.2, + "valid_targets_min": 1270 + }, + { + "epoch": 3.5611907386990076, + "grad_norm": 0.5614393447480441, + "learning_rate": 2.2876472362933333e-05, + "loss": 0.2368, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10511431097984314, + "step": 6460, + "valid_targets_mean": 3632.4, + "valid_targets_min": 584 + }, + { + "epoch": 3.563947078280044, + "grad_norm": 0.5558814427331218, + "learning_rate": 2.2849265151476016e-05, + "loss": 0.2324, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1263851523399353, + "step": 6465, + "valid_targets_mean": 3856.4, + "valid_targets_min": 1525 + }, + { + "epoch": 3.5667034178610804, + "grad_norm": 0.5710248844074187, + "learning_rate": 2.2822052556923145e-05, + "loss": 0.2176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11321952939033508, + "step": 6470, + "valid_targets_mean": 3541.8, + "valid_targets_min": 1120 + }, + { + "epoch": 3.5694597574421167, + "grad_norm": 0.5947906350746759, + "learning_rate": 2.279483463068728e-05, + "loss": 0.2209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14571425318717957, + "step": 6475, + "valid_targets_mean": 5234.5, + "valid_targets_min": 1423 + }, + { + "epoch": 3.572216097023153, + "grad_norm": 0.6979967198577909, + "learning_rate": 2.2767611424191043e-05, + "loss": 0.2155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12128198891878128, + "step": 6480, + "valid_targets_mean": 2462.5, + "valid_targets_min": 972 + }, + { + "epoch": 3.5749724366041895, + "grad_norm": 0.6056543548996169, + "learning_rate": 2.274038298886704e-05, + "loss": 0.2169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11020754277706146, + "step": 6485, + "valid_targets_mean": 3269.8, + "valid_targets_min": 1907 + }, + { + "epoch": 3.577728776185226, + "grad_norm": 0.6545775645632402, + "learning_rate": 2.271314937615775e-05, + "loss": 0.2111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12354698032140732, + "step": 6490, + "valid_targets_mean": 3808.8, + "valid_targets_min": 773 + }, + { + "epoch": 3.5804851157662623, + "grad_norm": 0.5879239617162749, + "learning_rate": 2.2685910637515427e-05, + "loss": 0.2136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09499567747116089, + "step": 6495, + "valid_targets_mean": 2875.2, + "valid_targets_min": 567 + }, + { + "epoch": 3.5832414553472987, + "grad_norm": 0.5802729702402881, + "learning_rate": 2.2658666824402037e-05, + "loss": 0.2155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08672034740447998, + "step": 6500, + "valid_targets_mean": 2599.8, + "valid_targets_min": 1236 + }, + { + "epoch": 3.585997794928335, + "grad_norm": 0.7053795730649799, + "learning_rate": 2.26314179882891e-05, + "loss": 0.2459, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10564807057380676, + "step": 6505, + "valid_targets_mean": 2615.4, + "valid_targets_min": 1217 + }, + { + "epoch": 3.5887541345093714, + "grad_norm": 0.6361202461100199, + "learning_rate": 2.2604164180657648e-05, + "loss": 0.2179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08934658765792847, + "step": 6510, + "valid_targets_mean": 2228.8, + "valid_targets_min": 1287 + }, + { + "epoch": 3.591510474090408, + "grad_norm": 0.6832036497372183, + "learning_rate": 2.2576905452998093e-05, + "loss": 0.2284, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10695388168096542, + "step": 6515, + "valid_targets_mean": 3568.4, + "valid_targets_min": 1731 + }, + { + "epoch": 3.594266813671444, + "grad_norm": 0.580404208131248, + "learning_rate": 2.2549641856810144e-05, + "loss": 0.2136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08078062534332275, + "step": 6520, + "valid_targets_mean": 2266.4, + "valid_targets_min": 755 + }, + { + "epoch": 3.5970231532524806, + "grad_norm": 0.615190865972318, + "learning_rate": 2.252237344360272e-05, + "loss": 0.2175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13265864551067352, + "step": 6525, + "valid_targets_mean": 4111.6, + "valid_targets_min": 2248 + }, + { + "epoch": 3.599779492833517, + "grad_norm": 0.9549194146630159, + "learning_rate": 2.249510026489383e-05, + "loss": 0.2169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09488487243652344, + "step": 6530, + "valid_targets_mean": 3105.5, + "valid_targets_min": 1272 + }, + { + "epoch": 3.6025358324145533, + "grad_norm": 0.5891339979261697, + "learning_rate": 2.2467822372210487e-05, + "loss": 0.2172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10120606422424316, + "step": 6535, + "valid_targets_mean": 3186.8, + "valid_targets_min": 582 + }, + { + "epoch": 3.6052921719955897, + "grad_norm": 0.5911831367081285, + "learning_rate": 2.244053981708861e-05, + "loss": 0.2231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15170781314373016, + "step": 6540, + "valid_targets_mean": 4445.9, + "valid_targets_min": 3153 + }, + { + "epoch": 3.608048511576626, + "grad_norm": 0.6008640619569721, + "learning_rate": 2.241325265107293e-05, + "loss": 0.2203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12522479891777039, + "step": 6545, + "valid_targets_mean": 4120.8, + "valid_targets_min": 2344 + }, + { + "epoch": 3.6108048511576625, + "grad_norm": 0.613182237700252, + "learning_rate": 2.238596092571689e-05, + "loss": 0.2115, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11675232648849487, + "step": 6550, + "valid_targets_mean": 3246.4, + "valid_targets_min": 1226 + }, + { + "epoch": 3.613561190738699, + "grad_norm": 0.5733801991628535, + "learning_rate": 2.2358664692582542e-05, + "loss": 0.2322, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11356987804174423, + "step": 6555, + "valid_targets_mean": 4125.2, + "valid_targets_min": 2852 + }, + { + "epoch": 3.6163175303197352, + "grad_norm": 0.5538285808602076, + "learning_rate": 2.2331364003240456e-05, + "loss": 0.2272, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10596924275159836, + "step": 6560, + "valid_targets_mean": 3805.8, + "valid_targets_min": 1767 + }, + { + "epoch": 3.6190738699007716, + "grad_norm": 0.5858972080363668, + "learning_rate": 2.230405890926962e-05, + "loss": 0.233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1469254046678543, + "step": 6565, + "valid_targets_mean": 4721.8, + "valid_targets_min": 3155 + }, + { + "epoch": 3.621830209481808, + "grad_norm": 0.6703875113512987, + "learning_rate": 2.2276749462257355e-05, + "loss": 0.2318, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11139967292547226, + "step": 6570, + "valid_targets_mean": 3387.1, + "valid_targets_min": 952 + }, + { + "epoch": 3.6245865490628444, + "grad_norm": 0.6073256430583297, + "learning_rate": 2.2249435713799188e-05, + "loss": 0.2304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12232047319412231, + "step": 6575, + "valid_targets_mean": 4148.1, + "valid_targets_min": 1209 + }, + { + "epoch": 3.6273428886438808, + "grad_norm": 0.6247375607385717, + "learning_rate": 2.2222117715498788e-05, + "loss": 0.2296, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10213878005743027, + "step": 6580, + "valid_targets_mean": 2986.8, + "valid_targets_min": 1012 + }, + { + "epoch": 3.630099228224917, + "grad_norm": 0.5965608329221027, + "learning_rate": 2.2194795518967847e-05, + "loss": 0.2334, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09825629740953445, + "step": 6585, + "valid_targets_mean": 3436.1, + "valid_targets_min": 1245 + }, + { + "epoch": 3.6328555678059535, + "grad_norm": 0.588709868386667, + "learning_rate": 2.2167469175825974e-05, + "loss": 0.2166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09571486711502075, + "step": 6590, + "valid_targets_mean": 3467.0, + "valid_targets_min": 467 + }, + { + "epoch": 3.63561190738699, + "grad_norm": 0.6431533238460158, + "learning_rate": 2.214013873770065e-05, + "loss": 0.2238, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10747207701206207, + "step": 6595, + "valid_targets_mean": 3260.0, + "valid_targets_min": 1555 + }, + { + "epoch": 3.6383682469680263, + "grad_norm": 0.6802909133302806, + "learning_rate": 2.211280425622705e-05, + "loss": 0.2304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11701985448598862, + "step": 6600, + "valid_targets_mean": 3240.5, + "valid_targets_min": 1007 + }, + { + "epoch": 3.6411245865490627, + "grad_norm": 0.5591751744916719, + "learning_rate": 2.2085465783048022e-05, + "loss": 0.2228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11814791709184647, + "step": 6605, + "valid_targets_mean": 4293.6, + "valid_targets_min": 1320 + }, + { + "epoch": 3.643880926130099, + "grad_norm": 0.5368127247124366, + "learning_rate": 2.2058123369813932e-05, + "loss": 0.2212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08829823136329651, + "step": 6610, + "valid_targets_mean": 3551.4, + "valid_targets_min": 2272 + }, + { + "epoch": 3.646637265711136, + "grad_norm": 0.6042186648772039, + "learning_rate": 2.203077706818259e-05, + "loss": 0.2115, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1014462411403656, + "step": 6615, + "valid_targets_mean": 2984.4, + "valid_targets_min": 1356 + }, + { + "epoch": 3.649393605292172, + "grad_norm": 0.6016873513543443, + "learning_rate": 2.2003426929819184e-05, + "loss": 0.2259, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12815852463245392, + "step": 6620, + "valid_targets_mean": 3562.2, + "valid_targets_min": 869 + }, + { + "epoch": 3.6521499448732087, + "grad_norm": 0.5718623864034521, + "learning_rate": 2.1976073006396113e-05, + "loss": 0.2026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10877338796854019, + "step": 6625, + "valid_targets_mean": 3966.9, + "valid_targets_min": 1418 + }, + { + "epoch": 3.6549062844542446, + "grad_norm": 0.5933523752492752, + "learning_rate": 2.1948715349592944e-05, + "loss": 0.2224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11256639659404755, + "step": 6630, + "valid_targets_mean": 4043.0, + "valid_targets_min": 2179 + }, + { + "epoch": 3.6576626240352814, + "grad_norm": 0.5732456465486543, + "learning_rate": 2.1921354011096302e-05, + "loss": 0.2187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08235008269548416, + "step": 6635, + "valid_targets_mean": 3022.6, + "valid_targets_min": 1268 + }, + { + "epoch": 3.6604189636163174, + "grad_norm": 0.6000094659000093, + "learning_rate": 2.1893989042599744e-05, + "loss": 0.2272, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11048313975334167, + "step": 6640, + "valid_targets_mean": 3557.9, + "valid_targets_min": 1239 + }, + { + "epoch": 3.663175303197354, + "grad_norm": 0.6187403767964837, + "learning_rate": 2.1866620495803722e-05, + "loss": 0.209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09335967898368835, + "step": 6645, + "valid_targets_mean": 3356.5, + "valid_targets_min": 1326 + }, + { + "epoch": 3.66593164277839, + "grad_norm": 0.6442768171066076, + "learning_rate": 2.183924842241542e-05, + "loss": 0.2326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14150092005729675, + "step": 6650, + "valid_targets_mean": 4695.0, + "valid_targets_min": 2280 + }, + { + "epoch": 3.668687982359427, + "grad_norm": 0.5722993454404113, + "learning_rate": 2.1811872874148688e-05, + "loss": 0.2188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11437917500734329, + "step": 6655, + "valid_targets_mean": 4493.5, + "valid_targets_min": 2296 + }, + { + "epoch": 3.671444321940463, + "grad_norm": 0.623642224044295, + "learning_rate": 2.178449390272396e-05, + "loss": 0.2141, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09953156113624573, + "step": 6660, + "valid_targets_mean": 2813.5, + "valid_targets_min": 806 + }, + { + "epoch": 3.6742006615214997, + "grad_norm": 0.6062894964544833, + "learning_rate": 2.175711155986811e-05, + "loss": 0.2088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.122981496155262, + "step": 6665, + "valid_targets_mean": 3877.6, + "valid_targets_min": 1462 + }, + { + "epoch": 3.6769570011025356, + "grad_norm": 0.6021069211676066, + "learning_rate": 2.172972589731441e-05, + "loss": 0.2127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10721150785684586, + "step": 6670, + "valid_targets_mean": 3635.1, + "valid_targets_min": 2350 + }, + { + "epoch": 3.6797133406835725, + "grad_norm": 0.9792099978347174, + "learning_rate": 2.170233696680238e-05, + "loss": 0.2081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1003856509923935, + "step": 6675, + "valid_targets_mean": 3803.1, + "valid_targets_min": 1357 + }, + { + "epoch": 3.6824696802646084, + "grad_norm": 0.5712806272293831, + "learning_rate": 2.167494482007772e-05, + "loss": 0.2039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11702825129032135, + "step": 6680, + "valid_targets_mean": 4294.6, + "valid_targets_min": 1054 + }, + { + "epoch": 3.6852260198456452, + "grad_norm": 0.696324388822135, + "learning_rate": 2.164754950889221e-05, + "loss": 0.2071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1182757094502449, + "step": 6685, + "valid_targets_mean": 3288.0, + "valid_targets_min": 1985 + }, + { + "epoch": 3.687982359426681, + "grad_norm": 0.5753789312691026, + "learning_rate": 2.1620151085003625e-05, + "loss": 0.2059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08945441246032715, + "step": 6690, + "valid_targets_mean": 3374.0, + "valid_targets_min": 1149 + }, + { + "epoch": 3.690738699007718, + "grad_norm": 0.603353494137245, + "learning_rate": 2.1592749600175594e-05, + "loss": 0.2225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09343340247869492, + "step": 6695, + "valid_targets_mean": 3209.8, + "valid_targets_min": 1091 + }, + { + "epoch": 3.693495038588754, + "grad_norm": 0.6063944605443239, + "learning_rate": 2.1565345106177533e-05, + "loss": 0.2315, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09886176884174347, + "step": 6700, + "valid_targets_mean": 3629.9, + "valid_targets_min": 1798 + }, + { + "epoch": 3.6962513781697908, + "grad_norm": 0.6109068272394444, + "learning_rate": 2.1537937654784556e-05, + "loss": 0.2066, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09575532376766205, + "step": 6705, + "valid_targets_mean": 2727.5, + "valid_targets_min": 1168 + }, + { + "epoch": 3.6990077177508267, + "grad_norm": 0.6286002422806978, + "learning_rate": 2.1510527297777357e-05, + "loss": 0.2128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10445833206176758, + "step": 6710, + "valid_targets_mean": 3118.4, + "valid_targets_min": 1429 + }, + { + "epoch": 3.7017640573318635, + "grad_norm": 0.6419595313360634, + "learning_rate": 2.148311408694212e-05, + "loss": 0.2135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08599317073822021, + "step": 6715, + "valid_targets_mean": 2699.0, + "valid_targets_min": 856 + }, + { + "epoch": 3.7045203969128995, + "grad_norm": 0.5501698589692832, + "learning_rate": 2.1455698074070422e-05, + "loss": 0.2149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09416841715574265, + "step": 6720, + "valid_targets_mean": 4466.4, + "valid_targets_min": 2686 + }, + { + "epoch": 3.7072767364939363, + "grad_norm": 0.5941879215362337, + "learning_rate": 2.1428279310959136e-05, + "loss": 0.2263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11734535545110703, + "step": 6725, + "valid_targets_mean": 3606.0, + "valid_targets_min": 1593 + }, + { + "epoch": 3.7100330760749722, + "grad_norm": 0.583786724893375, + "learning_rate": 2.1400857849410324e-05, + "loss": 0.2183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11613878607749939, + "step": 6730, + "valid_targets_mean": 4037.2, + "valid_targets_min": 958 + }, + { + "epoch": 3.712789415656009, + "grad_norm": 0.6118099904856805, + "learning_rate": 2.1373433741231144e-05, + "loss": 0.212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10636115074157715, + "step": 6735, + "valid_targets_mean": 3139.6, + "valid_targets_min": 978 + }, + { + "epoch": 3.715545755237045, + "grad_norm": 0.5940199375698697, + "learning_rate": 2.1346007038233774e-05, + "loss": 0.2219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09860540181398392, + "step": 6740, + "valid_targets_mean": 3547.2, + "valid_targets_min": 668 + }, + { + "epoch": 3.718302094818082, + "grad_norm": 0.625988725636132, + "learning_rate": 2.131857779223528e-05, + "loss": 0.2214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12598584592342377, + "step": 6745, + "valid_targets_mean": 3632.4, + "valid_targets_min": 507 + }, + { + "epoch": 3.7210584343991178, + "grad_norm": 0.39762273471556214, + "learning_rate": 2.129114605505752e-05, + "loss": 0.125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047671593725681305, + "step": 6750, + "valid_targets_mean": 2741.9, + "valid_targets_min": 455 + }, + { + "epoch": 3.7238147739801546, + "grad_norm": 0.3836891465115751, + "learning_rate": 2.126371187852708e-05, + "loss": 0.0889, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042357951402664185, + "step": 6755, + "valid_targets_mean": 3359.0, + "valid_targets_min": 2797 + }, + { + "epoch": 3.7265711135611905, + "grad_norm": 0.3679871610537133, + "learning_rate": 2.1236275314475144e-05, + "loss": 0.0849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049267180263996124, + "step": 6760, + "valid_targets_mean": 3299.6, + "valid_targets_min": 2059 + }, + { + "epoch": 3.7293274531422274, + "grad_norm": 0.4729512928763202, + "learning_rate": 2.120883641473742e-05, + "loss": 0.1026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050528425723314285, + "step": 6765, + "valid_targets_mean": 3373.4, + "valid_targets_min": 2197 + }, + { + "epoch": 3.7320837927232633, + "grad_norm": 0.47026375787138763, + "learning_rate": 2.1181395231154004e-05, + "loss": 0.0848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04918414726853371, + "step": 6770, + "valid_targets_mean": 2768.4, + "valid_targets_min": 1192 + }, + { + "epoch": 3.7348401323043, + "grad_norm": 0.6913344130357649, + "learning_rate": 2.1153951815569326e-05, + "loss": 0.0995, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06120513752102852, + "step": 6775, + "valid_targets_mean": 1055.5, + "valid_targets_min": 803 + }, + { + "epoch": 3.737596471885336, + "grad_norm": 0.45591809952951007, + "learning_rate": 2.112650621983203e-05, + "loss": 0.1405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05755969136953354, + "step": 6780, + "valid_targets_mean": 3310.8, + "valid_targets_min": 1943 + }, + { + "epoch": 3.740352811466373, + "grad_norm": 0.44261716993685163, + "learning_rate": 2.1099058495794874e-05, + "loss": 0.0927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061143867671489716, + "step": 6785, + "valid_targets_mean": 3068.4, + "valid_targets_min": 777 + }, + { + "epoch": 3.743109151047409, + "grad_norm": 0.3272195714489898, + "learning_rate": 2.1071608695314644e-05, + "loss": 0.106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03530782461166382, + "step": 6790, + "valid_targets_mean": 3493.8, + "valid_targets_min": 794 + }, + { + "epoch": 3.7458654906284456, + "grad_norm": 0.29502077155654666, + "learning_rate": 2.104415687025204e-05, + "loss": 0.1073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04011997953057289, + "step": 6795, + "valid_targets_mean": 4785.0, + "valid_targets_min": 2792 + }, + { + "epoch": 3.7486218302094816, + "grad_norm": 0.32460049310649847, + "learning_rate": 2.1016703072471593e-05, + "loss": 0.0829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027716509997844696, + "step": 6800, + "valid_targets_mean": 2383.8, + "valid_targets_min": 571 + }, + { + "epoch": 3.7513781697905184, + "grad_norm": 0.36143407283953316, + "learning_rate": 2.0989247353841557e-05, + "loss": 0.091, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04930020868778229, + "step": 6805, + "valid_targets_mean": 3945.8, + "valid_targets_min": 2633 + }, + { + "epoch": 3.7541345093715544, + "grad_norm": 0.5316974296586129, + "learning_rate": 2.0961789766233824e-05, + "loss": 0.0922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06081647425889969, + "step": 6810, + "valid_targets_mean": 1657.4, + "valid_targets_min": 590 + }, + { + "epoch": 3.756890848952591, + "grad_norm": 0.3854726037045277, + "learning_rate": 2.0934330361523812e-05, + "loss": 0.1226, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03862528130412102, + "step": 6815, + "valid_targets_mean": 3209.0, + "valid_targets_min": 1346 + }, + { + "epoch": 3.759647188533627, + "grad_norm": 0.5045496150438159, + "learning_rate": 2.0906869191590365e-05, + "loss": 0.0871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058705173432826996, + "step": 6820, + "valid_targets_mean": 2070.0, + "valid_targets_min": 813 + }, + { + "epoch": 3.762403528114664, + "grad_norm": 0.3636237062394931, + "learning_rate": 2.0879406308315668e-05, + "loss": 0.0834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034107062965631485, + "step": 6825, + "valid_targets_mean": 1848.0, + "valid_targets_min": 905 + }, + { + "epoch": 3.7651598676957, + "grad_norm": 0.3554319040509037, + "learning_rate": 2.0851941763585147e-05, + "loss": 0.0755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04194355383515358, + "step": 6830, + "valid_targets_mean": 3711.9, + "valid_targets_min": 2833 + }, + { + "epoch": 3.7679162072767367, + "grad_norm": 0.4468480173343079, + "learning_rate": 2.082447560928736e-05, + "loss": 0.0985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04598527401685715, + "step": 6835, + "valid_targets_mean": 2754.4, + "valid_targets_min": 468 + }, + { + "epoch": 3.7706725468577726, + "grad_norm": 0.43796600056379015, + "learning_rate": 2.079700789731391e-05, + "loss": 0.0976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056760065257549286, + "step": 6840, + "valid_targets_mean": 3055.2, + "valid_targets_min": 802 + }, + { + "epoch": 3.7734288864388095, + "grad_norm": 0.5132217224238966, + "learning_rate": 2.0769538679559343e-05, + "loss": 0.1013, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04666544497013092, + "step": 6845, + "valid_targets_mean": 2917.2, + "valid_targets_min": 471 + }, + { + "epoch": 3.7761852260198454, + "grad_norm": 0.4280776403299201, + "learning_rate": 2.074206800792105e-05, + "loss": 0.0994, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044018156826496124, + "step": 6850, + "valid_targets_mean": 3326.2, + "valid_targets_min": 636 + }, + { + "epoch": 3.7789415656008822, + "grad_norm": 0.3574984262148037, + "learning_rate": 2.071459593429916e-05, + "loss": 0.1236, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035675838589668274, + "step": 6855, + "valid_targets_mean": 3525.8, + "valid_targets_min": 898 + }, + { + "epoch": 3.781697905181918, + "grad_norm": 0.40675667975256813, + "learning_rate": 2.0687122510596462e-05, + "loss": 0.0848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04494604468345642, + "step": 6860, + "valid_targets_mean": 2980.0, + "valid_targets_min": 1829 + }, + { + "epoch": 3.784454244762955, + "grad_norm": 0.6839569088829538, + "learning_rate": 2.06596477887183e-05, + "loss": 0.1167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.086229607462883, + "step": 6865, + "valid_targets_mean": 1320.9, + "valid_targets_min": 752 + }, + { + "epoch": 3.787210584343991, + "grad_norm": 0.5793936829356531, + "learning_rate": 2.0632171820572454e-05, + "loss": 0.12, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08738426864147186, + "step": 6870, + "valid_targets_mean": 2208.9, + "valid_targets_min": 596 + }, + { + "epoch": 3.7899669239250278, + "grad_norm": 0.3631149871404351, + "learning_rate": 2.0604694658069065e-05, + "loss": 0.0894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03291577100753784, + "step": 6875, + "valid_targets_mean": 3024.1, + "valid_targets_min": 890 + }, + { + "epoch": 3.7927232635060637, + "grad_norm": 0.5952903772644363, + "learning_rate": 2.057721635312054e-05, + "loss": 0.194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1724621057510376, + "step": 6880, + "valid_targets_mean": 2450.4, + "valid_targets_min": 828 + }, + { + "epoch": 3.7954796030871005, + "grad_norm": 0.4211175282305943, + "learning_rate": 2.054973695764143e-05, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041970014572143555, + "step": 6885, + "valid_targets_mean": 2688.0, + "valid_targets_min": 752 + }, + { + "epoch": 3.7982359426681365, + "grad_norm": 0.3682763040643309, + "learning_rate": 2.0522256523548362e-05, + "loss": 0.1045, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040831390768289566, + "step": 6890, + "valid_targets_mean": 3327.1, + "valid_targets_min": 1960 + }, + { + "epoch": 3.8009922822491733, + "grad_norm": 0.6457284446598185, + "learning_rate": 2.049477510275991e-05, + "loss": 0.1638, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21535450220108032, + "step": 6895, + "valid_targets_mean": 2710.6, + "valid_targets_min": 1188 + }, + { + "epoch": 3.8037486218302092, + "grad_norm": 0.3117188333963592, + "learning_rate": 2.0467292747196516e-05, + "loss": 0.1163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03372468426823616, + "step": 6900, + "valid_targets_mean": 4566.2, + "valid_targets_min": 1693 + }, + { + "epoch": 3.806504961411246, + "grad_norm": 0.34531643595570105, + "learning_rate": 2.04398095087804e-05, + "loss": 0.0895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033125244081020355, + "step": 6905, + "valid_targets_mean": 3414.8, + "valid_targets_min": 785 + }, + { + "epoch": 3.809261300992282, + "grad_norm": 0.3809563091156694, + "learning_rate": 2.041232543943543e-05, + "loss": 0.0984, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05176117271184921, + "step": 6910, + "valid_targets_mean": 3680.8, + "valid_targets_min": 845 + }, + { + "epoch": 3.812017640573319, + "grad_norm": 0.32194250341122227, + "learning_rate": 2.038484059108707e-05, + "loss": 0.087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033377163112163544, + "step": 6915, + "valid_targets_mean": 2728.6, + "valid_targets_min": 485 + }, + { + "epoch": 3.8147739801543548, + "grad_norm": 0.4383398122111055, + "learning_rate": 2.0357355015662228e-05, + "loss": 0.0885, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056433141231536865, + "step": 6920, + "valid_targets_mean": 3904.0, + "valid_targets_min": 2429 + }, + { + "epoch": 3.8175303197353916, + "grad_norm": 0.47672075307976614, + "learning_rate": 2.0329868765089197e-05, + "loss": 0.0846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034156881272792816, + "step": 6925, + "valid_targets_mean": 1569.0, + "valid_targets_min": 723 + }, + { + "epoch": 3.8202866593164275, + "grad_norm": 0.42583815576892564, + "learning_rate": 2.0302381891297553e-05, + "loss": 0.0861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03923425450921059, + "step": 6930, + "valid_targets_mean": 3352.5, + "valid_targets_min": 1763 + }, + { + "epoch": 3.8230429988974644, + "grad_norm": 0.36430050368458755, + "learning_rate": 2.0274894446218042e-05, + "loss": 0.0771, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04590316116809845, + "step": 6935, + "valid_targets_mean": 3782.9, + "valid_targets_min": 2786 + }, + { + "epoch": 3.8257993384785003, + "grad_norm": 0.3557358400844802, + "learning_rate": 2.024740648178249e-05, + "loss": 0.1338, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03226230666041374, + "step": 6940, + "valid_targets_mean": 3618.4, + "valid_targets_min": 2915 + }, + { + "epoch": 3.828555678059537, + "grad_norm": 0.3586221995789157, + "learning_rate": 2.0219918049923707e-05, + "loss": 0.0679, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033270593732595444, + "step": 6945, + "valid_targets_mean": 3032.0, + "valid_targets_min": 718 + }, + { + "epoch": 3.831312017640573, + "grad_norm": 0.30401695741708973, + "learning_rate": 2.0192429202575372e-05, + "loss": 0.0785, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.024654366075992584, + "step": 6950, + "valid_targets_mean": 3490.0, + "valid_targets_min": 2531 + }, + { + "epoch": 3.83406835722161, + "grad_norm": 0.3951985010174385, + "learning_rate": 2.0164939991671976e-05, + "loss": 0.0812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03936031833291054, + "step": 6955, + "valid_targets_mean": 3218.5, + "valid_targets_min": 944 + }, + { + "epoch": 3.836824696802646, + "grad_norm": 0.478819037251662, + "learning_rate": 2.0137450469148674e-05, + "loss": 0.0947, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03858153522014618, + "step": 6960, + "valid_targets_mean": 1017.4, + "valid_targets_min": 487 + }, + { + "epoch": 3.8395810363836826, + "grad_norm": 0.34987309473671324, + "learning_rate": 2.0109960686941223e-05, + "loss": 0.1086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04548937454819679, + "step": 6965, + "valid_targets_mean": 3657.1, + "valid_targets_min": 2788 + }, + { + "epoch": 3.8423373759647186, + "grad_norm": 0.48739516198672544, + "learning_rate": 2.0082470696985856e-05, + "loss": 0.1228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07036057114601135, + "step": 6970, + "valid_targets_mean": 1813.9, + "valid_targets_min": 795 + }, + { + "epoch": 3.8450937155457554, + "grad_norm": 0.4467583757322096, + "learning_rate": 2.005498055121921e-05, + "loss": 0.0948, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044498175382614136, + "step": 6975, + "valid_targets_mean": 2663.5, + "valid_targets_min": 521 + }, + { + "epoch": 3.8478500551267913, + "grad_norm": 0.42250091539254586, + "learning_rate": 2.0027490301578217e-05, + "loss": 0.111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05675992742180824, + "step": 6980, + "valid_targets_mean": 3276.8, + "valid_targets_min": 707 + }, + { + "epoch": 3.850606394707828, + "grad_norm": 0.3503680240957169, + "learning_rate": 2e-05, + "loss": 0.1237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0356384702026844, + "step": 6985, + "valid_targets_mean": 2947.5, + "valid_targets_min": 2657 + }, + { + "epoch": 3.853362734288864, + "grad_norm": 0.3832849014075428, + "learning_rate": 1.997250969842179e-05, + "loss": 0.0945, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03920873627066612, + "step": 6990, + "valid_targets_mean": 3084.2, + "valid_targets_min": 885 + }, + { + "epoch": 3.856119073869901, + "grad_norm": 0.4898938491358198, + "learning_rate": 1.9945019448780798e-05, + "loss": 0.1124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03843805938959122, + "step": 6995, + "valid_targets_mean": 3782.1, + "valid_targets_min": 2384 + }, + { + "epoch": 3.8588754134509373, + "grad_norm": 0.40773695795926795, + "learning_rate": 1.9917529303014148e-05, + "loss": 0.0978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05270010232925415, + "step": 7000, + "valid_targets_mean": 3404.1, + "valid_targets_min": 1162 + }, + { + "epoch": 3.8616317530319737, + "grad_norm": 0.4094828306012488, + "learning_rate": 1.9890039313058784e-05, + "loss": 0.0898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05160021409392357, + "step": 7005, + "valid_targets_mean": 5079.5, + "valid_targets_min": 3681 + }, + { + "epoch": 3.86438809261301, + "grad_norm": 0.47509480179778235, + "learning_rate": 1.986254953085133e-05, + "loss": 0.0824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05090019106864929, + "step": 7010, + "valid_targets_mean": 2498.4, + "valid_targets_min": 592 + }, + { + "epoch": 3.8671444321940465, + "grad_norm": 0.2809793319420022, + "learning_rate": 1.9835060008328028e-05, + "loss": 0.0773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027998534962534904, + "step": 7015, + "valid_targets_mean": 4521.8, + "valid_targets_min": 3791 + }, + { + "epoch": 3.869900771775083, + "grad_norm": 0.3640895804206597, + "learning_rate": 1.9807570797424634e-05, + "loss": 0.0854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03761552274227142, + "step": 7020, + "valid_targets_mean": 3725.8, + "valid_targets_min": 2852 + }, + { + "epoch": 3.8726571113561192, + "grad_norm": 0.44256195431683293, + "learning_rate": 1.97800819500763e-05, + "loss": 0.1002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04798021540045738, + "step": 7025, + "valid_targets_mean": 1102.9, + "valid_targets_min": 753 + }, + { + "epoch": 3.8754134509371556, + "grad_norm": 0.4022499177624629, + "learning_rate": 1.9752593518217514e-05, + "loss": 0.0887, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046220533549785614, + "step": 7030, + "valid_targets_mean": 3868.9, + "valid_targets_min": 2434 + }, + { + "epoch": 3.878169790518192, + "grad_norm": 0.6660626836601903, + "learning_rate": 1.972510555378196e-05, + "loss": 0.1068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05930892750620842, + "step": 7035, + "valid_targets_mean": 1375.4, + "valid_targets_min": 780 + }, + { + "epoch": 3.8809261300992284, + "grad_norm": 0.3298472757749275, + "learning_rate": 1.969761810870245e-05, + "loss": 0.0718, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03998961299657822, + "step": 7040, + "valid_targets_mean": 3640.8, + "valid_targets_min": 900 + }, + { + "epoch": 3.8836824696802648, + "grad_norm": 0.41945669087888554, + "learning_rate": 1.967013123491081e-05, + "loss": 0.0916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05816015601158142, + "step": 7045, + "valid_targets_mean": 2962.4, + "valid_targets_min": 653 + }, + { + "epoch": 3.886438809261301, + "grad_norm": 0.375438535892087, + "learning_rate": 1.964264498433778e-05, + "loss": 0.0909, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04650282859802246, + "step": 7050, + "valid_targets_mean": 3711.2, + "valid_targets_min": 2758 + }, + { + "epoch": 3.8891951488423375, + "grad_norm": 0.5108333780362649, + "learning_rate": 1.9615159408912937e-05, + "loss": 0.0788, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05346187204122543, + "step": 7055, + "valid_targets_mean": 1395.0, + "valid_targets_min": 623 + }, + { + "epoch": 3.891951488423374, + "grad_norm": 0.4149125660882017, + "learning_rate": 1.9587674560564573e-05, + "loss": 0.0901, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04219534993171692, + "step": 7060, + "valid_targets_mean": 2325.1, + "valid_targets_min": 822 + }, + { + "epoch": 3.8947078280044103, + "grad_norm": 0.4393316987540876, + "learning_rate": 1.956019049121961e-05, + "loss": 0.0894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03755203261971474, + "step": 7065, + "valid_targets_mean": 1896.8, + "valid_targets_min": 935 + }, + { + "epoch": 3.8974641675854467, + "grad_norm": 0.3807169346469263, + "learning_rate": 1.9532707252803488e-05, + "loss": 0.1058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04622025787830353, + "step": 7070, + "valid_targets_mean": 3336.8, + "valid_targets_min": 1660 + }, + { + "epoch": 3.900220507166483, + "grad_norm": 0.38992924549135116, + "learning_rate": 1.9505224897240097e-05, + "loss": 0.0872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03570862486958504, + "step": 7075, + "valid_targets_mean": 2459.1, + "valid_targets_min": 894 + }, + { + "epoch": 3.9029768467475194, + "grad_norm": 0.6896350809825289, + "learning_rate": 1.9477743476451644e-05, + "loss": 0.1088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04022292420268059, + "step": 7080, + "valid_targets_mean": 2051.8, + "valid_targets_min": 755 + }, + { + "epoch": 3.905733186328556, + "grad_norm": 0.6051743688279897, + "learning_rate": 1.9450263042358574e-05, + "loss": 0.1029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07311645895242691, + "step": 7085, + "valid_targets_mean": 1594.8, + "valid_targets_min": 518 + }, + { + "epoch": 3.908489525909592, + "grad_norm": 0.48713540814905126, + "learning_rate": 1.942278364687947e-05, + "loss": 0.0914, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05672542378306389, + "step": 7090, + "valid_targets_mean": 3199.2, + "valid_targets_min": 1026 + }, + { + "epoch": 3.9112458654906286, + "grad_norm": 0.28856618975559445, + "learning_rate": 1.939530534193094e-05, + "loss": 0.0768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029305139556527138, + "step": 7095, + "valid_targets_mean": 3631.5, + "valid_targets_min": 2736 + }, + { + "epoch": 3.914002205071665, + "grad_norm": 0.3433788174344583, + "learning_rate": 1.9367828179427553e-05, + "loss": 0.0858, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03297518938779831, + "step": 7100, + "valid_targets_mean": 2821.0, + "valid_targets_min": 1681 + }, + { + "epoch": 3.9167585446527013, + "grad_norm": 0.39974949653702635, + "learning_rate": 1.9340352211281707e-05, + "loss": 0.1154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04454413056373596, + "step": 7105, + "valid_targets_mean": 3006.9, + "valid_targets_min": 929 + }, + { + "epoch": 3.9195148842337377, + "grad_norm": 0.3418928851460359, + "learning_rate": 1.931287748940354e-05, + "loss": 0.0721, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026177622377872467, + "step": 7110, + "valid_targets_mean": 3051.2, + "valid_targets_min": 734 + }, + { + "epoch": 3.922271223814774, + "grad_norm": 0.318885755277024, + "learning_rate": 1.9285404065700847e-05, + "loss": 0.0886, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0326494425535202, + "step": 7115, + "valid_targets_mean": 5296.9, + "valid_targets_min": 3274 + }, + { + "epoch": 3.9250275633958105, + "grad_norm": 0.2811524376242534, + "learning_rate": 1.9257931992078956e-05, + "loss": 0.0844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03546970337629318, + "step": 7120, + "valid_targets_mean": 4010.9, + "valid_targets_min": 989 + }, + { + "epoch": 3.927783902976847, + "grad_norm": 0.2623036589081945, + "learning_rate": 1.923046132044066e-05, + "loss": 0.0714, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02581711858510971, + "step": 7125, + "valid_targets_mean": 4373.0, + "valid_targets_min": 944 + }, + { + "epoch": 3.9305402425578833, + "grad_norm": 0.4070563056298574, + "learning_rate": 1.9202992102686096e-05, + "loss": 0.0947, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060628559440374374, + "step": 7130, + "valid_targets_mean": 4134.4, + "valid_targets_min": 2713 + }, + { + "epoch": 3.9332965821389196, + "grad_norm": 0.3530581342587192, + "learning_rate": 1.9175524390712645e-05, + "loss": 0.0795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029528118669986725, + "step": 7135, + "valid_targets_mean": 3617.2, + "valid_targets_min": 423 + }, + { + "epoch": 3.936052921719956, + "grad_norm": 0.34336024930972925, + "learning_rate": 1.914805823641486e-05, + "loss": 0.1094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034355323761701584, + "step": 7140, + "valid_targets_mean": 3218.5, + "valid_targets_min": 528 + }, + { + "epoch": 3.9388092613009924, + "grad_norm": 0.35826616546918894, + "learning_rate": 1.9120593691684335e-05, + "loss": 0.1083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04015689715743065, + "step": 7145, + "valid_targets_mean": 2979.4, + "valid_targets_min": 850 + }, + { + "epoch": 3.941565600882029, + "grad_norm": 0.34696581073421257, + "learning_rate": 1.909313080840964e-05, + "loss": 0.0681, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03895070031285286, + "step": 7150, + "valid_targets_mean": 3030.6, + "valid_targets_min": 718 + }, + { + "epoch": 3.944321940463065, + "grad_norm": 0.35760102482069367, + "learning_rate": 1.9065669638476195e-05, + "loss": 0.1102, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03599917143583298, + "step": 7155, + "valid_targets_mean": 3710.5, + "valid_targets_min": 1245 + }, + { + "epoch": 3.9470782800441016, + "grad_norm": 0.3771229737048134, + "learning_rate": 1.903821023376618e-05, + "loss": 0.0839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050358034670352936, + "step": 7160, + "valid_targets_mean": 3314.6, + "valid_targets_min": 1374 + }, + { + "epoch": 3.949834619625138, + "grad_norm": 0.46564303496910947, + "learning_rate": 1.9010752646158447e-05, + "loss": 0.0926, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03450053185224533, + "step": 7165, + "valid_targets_mean": 1037.5, + "valid_targets_min": 592 + }, + { + "epoch": 3.9525909592061743, + "grad_norm": 0.49811113246584343, + "learning_rate": 1.8983296927528413e-05, + "loss": 0.1038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04449811577796936, + "step": 7170, + "valid_targets_mean": 1287.6, + "valid_targets_min": 517 + }, + { + "epoch": 3.9553472987872107, + "grad_norm": 0.38153956069893175, + "learning_rate": 1.8955843129747965e-05, + "loss": 0.1148, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03562698885798454, + "step": 7175, + "valid_targets_mean": 3389.4, + "valid_targets_min": 863 + }, + { + "epoch": 3.958103638368247, + "grad_norm": 0.329282149470936, + "learning_rate": 1.8928391304685363e-05, + "loss": 0.09, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032044414430856705, + "step": 7180, + "valid_targets_mean": 3737.6, + "valid_targets_min": 2999 + }, + { + "epoch": 3.9608599779492835, + "grad_norm": 0.40352754618111647, + "learning_rate": 1.8900941504205133e-05, + "loss": 0.0915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04727362096309662, + "step": 7185, + "valid_targets_mean": 3086.1, + "valid_targets_min": 672 + }, + { + "epoch": 3.96361631753032, + "grad_norm": 0.4399804430060452, + "learning_rate": 1.8873493780167972e-05, + "loss": 0.086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03557152673602104, + "step": 7190, + "valid_targets_mean": 1813.5, + "valid_targets_min": 613 + }, + { + "epoch": 3.9663726571113562, + "grad_norm": 0.42030014918435377, + "learning_rate": 1.8846048184430677e-05, + "loss": 0.0842, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04084409028291702, + "step": 7195, + "valid_targets_mean": 2684.0, + "valid_targets_min": 538 + }, + { + "epoch": 3.9691289966923926, + "grad_norm": 0.6504528986195015, + "learning_rate": 1.8818604768846003e-05, + "loss": 0.0795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0539637990295887, + "step": 7200, + "valid_targets_mean": 986.4, + "valid_targets_min": 754 + }, + { + "epoch": 3.971885336273429, + "grad_norm": 0.39875713441663585, + "learning_rate": 1.8791163585262588e-05, + "loss": 0.0974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039606183767318726, + "step": 7205, + "valid_targets_mean": 2936.8, + "valid_targets_min": 1095 + }, + { + "epoch": 3.9746416758544654, + "grad_norm": 0.44804268606676645, + "learning_rate": 1.876372468552486e-05, + "loss": 0.079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04469123110175133, + "step": 7210, + "valid_targets_mean": 3170.8, + "valid_targets_min": 734 + }, + { + "epoch": 3.9773980154355018, + "grad_norm": 0.34906639110337156, + "learning_rate": 1.8736288121472927e-05, + "loss": 0.0819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042665328830480576, + "step": 7215, + "valid_targets_mean": 3209.4, + "valid_targets_min": 1000 + }, + { + "epoch": 3.980154355016538, + "grad_norm": 0.40033188517250246, + "learning_rate": 1.8708853944942486e-05, + "loss": 0.0952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04932411015033722, + "step": 7220, + "valid_targets_mean": 4100.1, + "valid_targets_min": 2762 + }, + { + "epoch": 3.9829106945975745, + "grad_norm": 0.25144338099592467, + "learning_rate": 1.868142220776473e-05, + "loss": 0.0753, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028727633878588676, + "step": 7225, + "valid_targets_mean": 5708.9, + "valid_targets_min": 3667 + }, + { + "epoch": 3.985667034178611, + "grad_norm": 0.401975964931408, + "learning_rate": 1.865399296176623e-05, + "loss": 0.0805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04704137146472931, + "step": 7230, + "valid_targets_mean": 2413.6, + "valid_targets_min": 671 + }, + { + "epoch": 3.9884233737596473, + "grad_norm": 0.32639940943517975, + "learning_rate": 1.862656625876886e-05, + "loss": 0.0779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0324050635099411, + "step": 7235, + "valid_targets_mean": 3936.2, + "valid_targets_min": 790 + }, + { + "epoch": 3.9911797133406837, + "grad_norm": 0.48479043660093984, + "learning_rate": 1.8599142150589682e-05, + "loss": 0.1778, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08632122725248337, + "step": 7240, + "valid_targets_mean": 3071.2, + "valid_targets_min": 1054 + }, + { + "epoch": 3.99393605292172, + "grad_norm": 0.3378875181429133, + "learning_rate": 1.8571720689040867e-05, + "loss": 0.0785, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03829573094844818, + "step": 7245, + "valid_targets_mean": 3072.6, + "valid_targets_min": 980 + }, + { + "epoch": 3.9966923925027564, + "grad_norm": 0.395094617038262, + "learning_rate": 1.854430192592958e-05, + "loss": 0.0906, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0488162487745285, + "step": 7250, + "valid_targets_mean": 2980.1, + "valid_targets_min": 874 + }, + { + "epoch": 3.999448732083793, + "grad_norm": 0.5810192449336612, + "learning_rate": 1.8516885913057886e-05, + "loss": 0.1084, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11428497731685638, + "step": 7255, + "valid_targets_mean": 2231.4, + "valid_targets_min": 1272 + }, + { + "epoch": 4.002205071664829, + "grad_norm": 0.6111462471403595, + "learning_rate": 1.848947270222265e-05, + "loss": 0.1779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1526864469051361, + "step": 7260, + "valid_targets_mean": 8001.9, + "valid_targets_min": 6110 + }, + { + "epoch": 4.004961411245866, + "grad_norm": 0.40820476964928515, + "learning_rate": 1.8462062345215447e-05, + "loss": 0.1933, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09133513271808624, + "step": 7265, + "valid_targets_mean": 6637.6, + "valid_targets_min": 5446 + }, + { + "epoch": 4.0077177508269015, + "grad_norm": 0.3903541994133546, + "learning_rate": 1.8434654893822474e-05, + "loss": 0.183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09643463790416718, + "step": 7270, + "valid_targets_mean": 7098.2, + "valid_targets_min": 5340 + }, + { + "epoch": 4.010474090407938, + "grad_norm": 0.34741764402572006, + "learning_rate": 1.8407250399824416e-05, + "loss": 0.1774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08616428077220917, + "step": 7275, + "valid_targets_mean": 7605.0, + "valid_targets_min": 5582 + }, + { + "epoch": 4.013230429988974, + "grad_norm": 0.3435108706336901, + "learning_rate": 1.837984891499638e-05, + "loss": 0.1816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08527673035860062, + "step": 7280, + "valid_targets_mean": 8599.5, + "valid_targets_min": 6330 + }, + { + "epoch": 4.015986769570011, + "grad_norm": 0.38669099386841055, + "learning_rate": 1.8352450491107793e-05, + "loss": 0.1854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09242553263902664, + "step": 7285, + "valid_targets_mean": 7354.9, + "valid_targets_min": 5812 + }, + { + "epoch": 4.018743109151047, + "grad_norm": 0.36812774523526764, + "learning_rate": 1.8325055179922285e-05, + "loss": 0.1931, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10579974949359894, + "step": 7290, + "valid_targets_mean": 7989.4, + "valid_targets_min": 5251 + }, + { + "epoch": 4.021499448732084, + "grad_norm": 0.3476617291650182, + "learning_rate": 1.829766303319763e-05, + "loss": 0.1815, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07803281396627426, + "step": 7295, + "valid_targets_mean": 7022.4, + "valid_targets_min": 4116 + }, + { + "epoch": 4.02425578831312, + "grad_norm": 0.3806986785290795, + "learning_rate": 1.82702741026856e-05, + "loss": 0.1768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0751921609044075, + "step": 7300, + "valid_targets_mean": 5345.5, + "valid_targets_min": 3897 + }, + { + "epoch": 4.027012127894157, + "grad_norm": 0.37970654650989205, + "learning_rate": 1.8242888440131894e-05, + "loss": 0.1642, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07701321691274643, + "step": 7305, + "valid_targets_mean": 6035.9, + "valid_targets_min": 4843 + }, + { + "epoch": 4.029768467475193, + "grad_norm": 0.4262804200289911, + "learning_rate": 1.8215506097276047e-05, + "loss": 0.1824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09059180319309235, + "step": 7310, + "valid_targets_mean": 6969.2, + "valid_targets_min": 5988 + }, + { + "epoch": 4.032524807056229, + "grad_norm": 0.3706564673057137, + "learning_rate": 1.8188127125851315e-05, + "loss": 0.1707, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08572712540626526, + "step": 7315, + "valid_targets_mean": 6150.4, + "valid_targets_min": 4637 + }, + { + "epoch": 4.035281146637265, + "grad_norm": 0.3611580446840618, + "learning_rate": 1.8160751577584587e-05, + "loss": 0.1776, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07902003824710846, + "step": 7320, + "valid_targets_mean": 6220.1, + "valid_targets_min": 4868 + }, + { + "epoch": 4.038037486218302, + "grad_norm": 0.355803607236488, + "learning_rate": 1.8133379504196288e-05, + "loss": 0.1697, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07790818810462952, + "step": 7325, + "valid_targets_mean": 7126.4, + "valid_targets_min": 5291 + }, + { + "epoch": 4.040793825799338, + "grad_norm": 0.3648699765290228, + "learning_rate": 1.8106010957400263e-05, + "loss": 0.1737, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08899436891078949, + "step": 7330, + "valid_targets_mean": 7184.9, + "valid_targets_min": 4699 + }, + { + "epoch": 4.043550165380375, + "grad_norm": 0.3657115045602958, + "learning_rate": 1.807864598890371e-05, + "loss": 0.178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08605834096670151, + "step": 7335, + "valid_targets_mean": 6518.6, + "valid_targets_min": 5707 + }, + { + "epoch": 4.046306504961411, + "grad_norm": 0.37363423552041697, + "learning_rate": 1.805128465040706e-05, + "loss": 0.1705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.087971031665802, + "step": 7340, + "valid_targets_mean": 6694.6, + "valid_targets_min": 4508 + }, + { + "epoch": 4.049062844542448, + "grad_norm": 0.4654071698598903, + "learning_rate": 1.8023926993603894e-05, + "loss": 0.178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09602256119251251, + "step": 7345, + "valid_targets_mean": 4555.8, + "valid_targets_min": 1033 + }, + { + "epoch": 4.051819184123484, + "grad_norm": 0.3920863828818436, + "learning_rate": 1.7996573070180823e-05, + "loss": 0.1911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08983595669269562, + "step": 7350, + "valid_targets_mean": 6398.8, + "valid_targets_min": 5380 + }, + { + "epoch": 4.0545755237045205, + "grad_norm": 0.6535505165876162, + "learning_rate": 1.7969222931817412e-05, + "loss": 0.1716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027201753109693527, + "step": 7355, + "valid_targets_mean": 579.6, + "valid_targets_min": 157 + }, + { + "epoch": 4.057331863285556, + "grad_norm": 0.36456776101342175, + "learning_rate": 1.7941876630186078e-05, + "loss": 0.1612, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09078461676836014, + "step": 7360, + "valid_targets_mean": 8202.1, + "valid_targets_min": 5037 + }, + { + "epoch": 4.060088202866593, + "grad_norm": 0.4099869538044665, + "learning_rate": 1.7914534216951988e-05, + "loss": 0.1746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09741708636283875, + "step": 7365, + "valid_targets_mean": 7208.6, + "valid_targets_min": 5089 + }, + { + "epoch": 4.062844542447629, + "grad_norm": 0.3895267810455791, + "learning_rate": 1.7887195743772953e-05, + "loss": 0.1712, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07707106322050095, + "step": 7370, + "valid_targets_mean": 6292.2, + "valid_targets_min": 4835 + }, + { + "epoch": 4.065600882028666, + "grad_norm": 0.36539527036558883, + "learning_rate": 1.785986126229936e-05, + "loss": 0.179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08723382651805878, + "step": 7375, + "valid_targets_mean": 7856.0, + "valid_targets_min": 5974 + }, + { + "epoch": 4.068357221609702, + "grad_norm": 0.34851622907032664, + "learning_rate": 1.783253082417403e-05, + "loss": 0.1699, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07936736196279526, + "step": 7380, + "valid_targets_mean": 7059.6, + "valid_targets_min": 4885 + }, + { + "epoch": 4.071113561190739, + "grad_norm": 0.394972104776565, + "learning_rate": 1.780520448103216e-05, + "loss": 0.1642, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08000268787145615, + "step": 7385, + "valid_targets_mean": 7414.6, + "valid_targets_min": 4181 + }, + { + "epoch": 4.073869900771775, + "grad_norm": 0.3633044576376274, + "learning_rate": 1.7777882284501215e-05, + "loss": 0.1505, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07970091700553894, + "step": 7390, + "valid_targets_mean": 7876.2, + "valid_targets_min": 5037 + }, + { + "epoch": 4.0766262403528115, + "grad_norm": 0.34208706155639407, + "learning_rate": 1.7750564286200816e-05, + "loss": 0.1545, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07256226241588593, + "step": 7395, + "valid_targets_mean": 8526.2, + "valid_targets_min": 6299 + }, + { + "epoch": 4.0793825799338475, + "grad_norm": 0.42756287546049543, + "learning_rate": 1.772325053774265e-05, + "loss": 0.1782, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08196178078651428, + "step": 7400, + "valid_targets_mean": 4955.2, + "valid_targets_min": 2831 + }, + { + "epoch": 4.082138919514884, + "grad_norm": 0.4198716780624949, + "learning_rate": 1.7695941090730385e-05, + "loss": 0.1603, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09520946443080902, + "step": 7405, + "valid_targets_mean": 7099.0, + "valid_targets_min": 5303 + }, + { + "epoch": 4.08489525909592, + "grad_norm": 0.431460234040602, + "learning_rate": 1.7668635996759547e-05, + "loss": 0.1816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09261700510978699, + "step": 7410, + "valid_targets_mean": 6837.5, + "valid_targets_min": 4678 + }, + { + "epoch": 4.087651598676957, + "grad_norm": 0.39066526056270684, + "learning_rate": 1.7641335307417465e-05, + "loss": 0.1866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0862879604101181, + "step": 7415, + "valid_targets_mean": 6603.6, + "valid_targets_min": 4604 + }, + { + "epoch": 4.090407938257993, + "grad_norm": 0.3743879484716084, + "learning_rate": 1.7614039074283117e-05, + "loss": 0.1803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0811166912317276, + "step": 7420, + "valid_targets_mean": 6302.1, + "valid_targets_min": 5395 + }, + { + "epoch": 4.09316427783903, + "grad_norm": 0.4019716530865773, + "learning_rate": 1.7586747348927075e-05, + "loss": 0.1735, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.082613006234169, + "step": 7425, + "valid_targets_mean": 6784.9, + "valid_targets_min": 4932 + }, + { + "epoch": 4.095920617420066, + "grad_norm": 0.41660841495067485, + "learning_rate": 1.7559460182911396e-05, + "loss": 0.1827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09621358662843704, + "step": 7430, + "valid_targets_mean": 6887.4, + "valid_targets_min": 5304 + }, + { + "epoch": 4.098676957001103, + "grad_norm": 0.38375174870289663, + "learning_rate": 1.7532177627789516e-05, + "loss": 0.1734, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08840475231409073, + "step": 7435, + "valid_targets_mean": 6433.5, + "valid_targets_min": 5161 + }, + { + "epoch": 4.1014332965821385, + "grad_norm": 0.39580804906790373, + "learning_rate": 1.7504899735106173e-05, + "loss": 0.1774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09132297337055206, + "step": 7440, + "valid_targets_mean": 7024.6, + "valid_targets_min": 4881 + }, + { + "epoch": 4.104189636163175, + "grad_norm": 0.45347654101309537, + "learning_rate": 1.7477626556397283e-05, + "loss": 0.18, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09860581159591675, + "step": 7445, + "valid_targets_mean": 7438.8, + "valid_targets_min": 5396 + }, + { + "epoch": 4.106945975744211, + "grad_norm": 0.39408047512296585, + "learning_rate": 1.745035814318986e-05, + "loss": 0.1805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09305399656295776, + "step": 7450, + "valid_targets_mean": 6787.4, + "valid_targets_min": 5254 + }, + { + "epoch": 4.109702315325248, + "grad_norm": 0.38198095199019894, + "learning_rate": 1.7423094547001913e-05, + "loss": 0.1785, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09418555349111557, + "step": 7455, + "valid_targets_mean": 7077.1, + "valid_targets_min": 5169 + }, + { + "epoch": 4.112458654906284, + "grad_norm": 0.39062623671661006, + "learning_rate": 1.7395835819342355e-05, + "loss": 0.1755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08522405475378036, + "step": 7460, + "valid_targets_mean": 5610.6, + "valid_targets_min": 4806 + }, + { + "epoch": 4.115214994487321, + "grad_norm": 0.8225194047106287, + "learning_rate": 1.7368582011710904e-05, + "loss": 0.1433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14413639903068542, + "step": 7465, + "valid_targets_mean": 2044.8, + "valid_targets_min": 133 + }, + { + "epoch": 4.117971334068357, + "grad_norm": 0.41766548602540865, + "learning_rate": 1.734133317559797e-05, + "loss": 0.178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08136571198701859, + "step": 7470, + "valid_targets_mean": 6655.9, + "valid_targets_min": 5224 + }, + { + "epoch": 4.120727673649394, + "grad_norm": 0.3940714681840455, + "learning_rate": 1.731408936248458e-05, + "loss": 0.172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09288594126701355, + "step": 7475, + "valid_targets_mean": 6184.5, + "valid_targets_min": 5509 + }, + { + "epoch": 4.12348401323043, + "grad_norm": 0.4404711488565934, + "learning_rate": 1.7286850623842258e-05, + "loss": 0.1764, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08836288750171661, + "step": 7480, + "valid_targets_mean": 6454.5, + "valid_targets_min": 5499 + }, + { + "epoch": 4.126240352811466, + "grad_norm": 0.3783672808484414, + "learning_rate": 1.7259617011132967e-05, + "loss": 0.1719, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08565817773342133, + "step": 7485, + "valid_targets_mean": 5767.6, + "valid_targets_min": 3887 + }, + { + "epoch": 4.128996692392502, + "grad_norm": 0.3995050503065588, + "learning_rate": 1.7232388575808964e-05, + "loss": 0.1709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08666878938674927, + "step": 7490, + "valid_targets_mean": 6327.5, + "valid_targets_min": 5163 + }, + { + "epoch": 4.131753031973539, + "grad_norm": 0.4253596553696065, + "learning_rate": 1.7205165369312728e-05, + "loss": 0.1789, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08881422877311707, + "step": 7495, + "valid_targets_mean": 5777.6, + "valid_targets_min": 4927 + }, + { + "epoch": 4.134509371554575, + "grad_norm": 0.5199429030518368, + "learning_rate": 1.7177947443076862e-05, + "loss": 0.1743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057098638266325, + "step": 7500, + "valid_targets_mean": 2577.2, + "valid_targets_min": 836 + }, + { + "epoch": 4.137265711135612, + "grad_norm": 0.43013528472106094, + "learning_rate": 1.715073484852399e-05, + "loss": 0.1448, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08291628956794739, + "step": 7505, + "valid_targets_mean": 5975.9, + "valid_targets_min": 4737 + }, + { + "epoch": 4.140022050716648, + "grad_norm": 0.419706256187936, + "learning_rate": 1.7123527637066673e-05, + "loss": 0.1705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0853375568985939, + "step": 7510, + "valid_targets_mean": 6735.0, + "valid_targets_min": 5014 + }, + { + "epoch": 4.142778390297685, + "grad_norm": 0.4286605668445149, + "learning_rate": 1.7096325860107286e-05, + "loss": 0.176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09100188314914703, + "step": 7515, + "valid_targets_mean": 6013.6, + "valid_targets_min": 5150 + }, + { + "epoch": 4.145534729878721, + "grad_norm": 0.35757440142300834, + "learning_rate": 1.7069129569037955e-05, + "loss": 0.1717, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08592434227466583, + "step": 7520, + "valid_targets_mean": 7069.6, + "valid_targets_min": 5334 + }, + { + "epoch": 4.1482910694597575, + "grad_norm": 0.37974338446214867, + "learning_rate": 1.704193881524042e-05, + "loss": 0.1671, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08809372782707214, + "step": 7525, + "valid_targets_mean": 7081.1, + "valid_targets_min": 5160 + }, + { + "epoch": 4.151047409040794, + "grad_norm": 0.41977796018020835, + "learning_rate": 1.7014753650085972e-05, + "loss": 0.1724, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08525887131690979, + "step": 7530, + "valid_targets_mean": 5478.6, + "valid_targets_min": 4764 + }, + { + "epoch": 4.15380374862183, + "grad_norm": 0.8946072362985246, + "learning_rate": 1.6987574124935353e-05, + "loss": 0.1704, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09000188112258911, + "step": 7535, + "valid_targets_mean": 1376.5, + "valid_targets_min": 597 + }, + { + "epoch": 4.156560088202866, + "grad_norm": 0.6975583245312854, + "learning_rate": 1.6960400291138625e-05, + "loss": 0.1621, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06978637725114822, + "step": 7540, + "valid_targets_mean": 1103.2, + "valid_targets_min": 467 + }, + { + "epoch": 4.159316427783903, + "grad_norm": 0.8036297014591007, + "learning_rate": 1.693323220003512e-05, + "loss": 0.1516, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07923336327075958, + "step": 7545, + "valid_targets_mean": 1417.8, + "valid_targets_min": 794 + }, + { + "epoch": 4.16207276736494, + "grad_norm": 0.9444604098519045, + "learning_rate": 1.690606990295331e-05, + "loss": 0.1489, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07401558011770248, + "step": 7550, + "valid_targets_mean": 1128.6, + "valid_targets_min": 653 + }, + { + "epoch": 4.164829106945976, + "grad_norm": 0.7771155364199565, + "learning_rate": 1.6878913451210715e-05, + "loss": 0.139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08810308575630188, + "step": 7555, + "valid_targets_mean": 1572.0, + "valid_targets_min": 608 + }, + { + "epoch": 4.167585446527012, + "grad_norm": 0.7966094483375156, + "learning_rate": 1.6851762896113827e-05, + "loss": 0.1431, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07523553818464279, + "step": 7560, + "valid_targets_mean": 1131.9, + "valid_targets_min": 677 + }, + { + "epoch": 4.1703417861080485, + "grad_norm": 0.8682326530155288, + "learning_rate": 1.682461828895799e-05, + "loss": 0.1508, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0838390439748764, + "step": 7565, + "valid_targets_mean": 1367.2, + "valid_targets_min": 912 + }, + { + "epoch": 4.173098125689085, + "grad_norm": 0.8787825619544478, + "learning_rate": 1.67974796810273e-05, + "loss": 0.1344, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06486639380455017, + "step": 7570, + "valid_targets_mean": 1098.5, + "valid_targets_min": 473 + }, + { + "epoch": 4.175854465270121, + "grad_norm": 0.7768724782323515, + "learning_rate": 1.6770347123594527e-05, + "loss": 0.1607, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0882253423333168, + "step": 7575, + "valid_targets_mean": 1426.1, + "valid_targets_min": 545 + }, + { + "epoch": 4.178610804851157, + "grad_norm": 0.7493436460392822, + "learning_rate": 1.674322066792102e-05, + "loss": 0.1366, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06206889823079109, + "step": 7580, + "valid_targets_mean": 1264.2, + "valid_targets_min": 707 + }, + { + "epoch": 4.181367144432194, + "grad_norm": 0.796308317213709, + "learning_rate": 1.6716100365256593e-05, + "loss": 0.1458, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0667402595281601, + "step": 7585, + "valid_targets_mean": 1255.5, + "valid_targets_min": 570 + }, + { + "epoch": 4.184123484013231, + "grad_norm": 0.7735399107466074, + "learning_rate": 1.668898626683942e-05, + "loss": 0.1421, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07857048511505127, + "step": 7590, + "valid_targets_mean": 1372.5, + "valid_targets_min": 504 + }, + { + "epoch": 4.186879823594267, + "grad_norm": 0.7686234909673326, + "learning_rate": 1.6661878423895975e-05, + "loss": 0.1421, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07009150832891464, + "step": 7595, + "valid_targets_mean": 1318.5, + "valid_targets_min": 903 + }, + { + "epoch": 4.189636163175303, + "grad_norm": 0.7203329191406345, + "learning_rate": 1.6634776887640902e-05, + "loss": 0.1319, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059526171535253525, + "step": 7600, + "valid_targets_mean": 1239.0, + "valid_targets_min": 531 + }, + { + "epoch": 4.19239250275634, + "grad_norm": 0.7780242158497488, + "learning_rate": 1.6607681709276938e-05, + "loss": 0.1277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06353896856307983, + "step": 7605, + "valid_targets_mean": 1423.9, + "valid_targets_min": 564 + }, + { + "epoch": 4.195148842337376, + "grad_norm": 0.841864937827762, + "learning_rate": 1.6580592939994796e-05, + "loss": 0.1345, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09444572031497955, + "step": 7610, + "valid_targets_mean": 1715.9, + "valid_targets_min": 835 + }, + { + "epoch": 4.197905181918412, + "grad_norm": 0.7714427945892655, + "learning_rate": 1.655351063097309e-05, + "loss": 0.1336, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05457676574587822, + "step": 7615, + "valid_targets_mean": 1033.8, + "valid_targets_min": 540 + }, + { + "epoch": 4.200661521499449, + "grad_norm": 0.7631545604180436, + "learning_rate": 1.6526434833378232e-05, + "loss": 0.1409, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06786838173866272, + "step": 7620, + "valid_targets_mean": 1251.9, + "valid_targets_min": 555 + }, + { + "epoch": 4.203417861080485, + "grad_norm": 0.6974080361959358, + "learning_rate": 1.649936559836431e-05, + "loss": 0.1389, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07348615676164627, + "step": 7625, + "valid_targets_mean": 1439.4, + "valid_targets_min": 723 + }, + { + "epoch": 4.206174200661522, + "grad_norm": 0.8666631082367596, + "learning_rate": 1.6472302977073044e-05, + "loss": 0.1367, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07675230503082275, + "step": 7630, + "valid_targets_mean": 1356.9, + "valid_targets_min": 509 + }, + { + "epoch": 4.208930540242558, + "grad_norm": 0.7421277272916458, + "learning_rate": 1.644524702063364e-05, + "loss": 0.1437, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061841633170843124, + "step": 7635, + "valid_targets_mean": 1469.1, + "valid_targets_min": 579 + }, + { + "epoch": 4.211686879823595, + "grad_norm": 0.7793594422578487, + "learning_rate": 1.6418197780162718e-05, + "loss": 0.1427, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06002141907811165, + "step": 7640, + "valid_targets_mean": 1372.0, + "valid_targets_min": 890 + }, + { + "epoch": 4.214443219404631, + "grad_norm": 0.7973910789193484, + "learning_rate": 1.63911553067642e-05, + "loss": 0.1367, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0833987221121788, + "step": 7645, + "valid_targets_mean": 1351.5, + "valid_targets_min": 558 + }, + { + "epoch": 4.2171995589856675, + "grad_norm": 0.7604919907865819, + "learning_rate": 1.6364119651529232e-05, + "loss": 0.1313, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060986027121543884, + "step": 7650, + "valid_targets_mean": 1269.9, + "valid_targets_min": 583 + }, + { + "epoch": 4.219955898566703, + "grad_norm": 0.8597070886974224, + "learning_rate": 1.633709086553609e-05, + "loss": 0.1341, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07764178514480591, + "step": 7655, + "valid_targets_mean": 1526.0, + "valid_targets_min": 486 + }, + { + "epoch": 4.22271223814774, + "grad_norm": 0.9329411020135167, + "learning_rate": 1.6310068999850045e-05, + "loss": 0.1385, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08613885939121246, + "step": 7660, + "valid_targets_mean": 1292.6, + "valid_targets_min": 667 + }, + { + "epoch": 4.225468577728776, + "grad_norm": 0.7455882101240536, + "learning_rate": 1.6283054105523315e-05, + "loss": 0.1424, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08118532598018646, + "step": 7665, + "valid_targets_mean": 1779.5, + "valid_targets_min": 920 + }, + { + "epoch": 4.228224917309813, + "grad_norm": 0.7400817019335751, + "learning_rate": 1.6256046233594937e-05, + "loss": 0.1399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08793629705905914, + "step": 7670, + "valid_targets_mean": 1883.9, + "valid_targets_min": 991 + }, + { + "epoch": 4.230981256890849, + "grad_norm": 0.7857464420298824, + "learning_rate": 1.622904543509068e-05, + "loss": 0.135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046498630195856094, + "step": 7675, + "valid_targets_mean": 869.4, + "valid_targets_min": 488 + }, + { + "epoch": 4.233737596471886, + "grad_norm": 0.7363493378292493, + "learning_rate": 1.620205176102296e-05, + "loss": 0.1344, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06506625562906265, + "step": 7680, + "valid_targets_mean": 1462.2, + "valid_targets_min": 584 + }, + { + "epoch": 4.236493936052922, + "grad_norm": 0.8009083824461161, + "learning_rate": 1.6175065262390724e-05, + "loss": 0.13, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059784185141325, + "step": 7685, + "valid_targets_mean": 1138.4, + "valid_targets_min": 626 + }, + { + "epoch": 4.2392502756339585, + "grad_norm": 0.8185054216476474, + "learning_rate": 1.614808599017936e-05, + "loss": 0.1402, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04990024119615555, + "step": 7690, + "valid_targets_mean": 941.1, + "valid_targets_min": 582 + }, + { + "epoch": 4.2420066152149944, + "grad_norm": 0.7917449080197263, + "learning_rate": 1.612111399536061e-05, + "loss": 0.1338, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07585994154214859, + "step": 7695, + "valid_targets_mean": 1557.1, + "valid_targets_min": 584 + }, + { + "epoch": 4.244762954796031, + "grad_norm": 0.9078681840697873, + "learning_rate": 1.609414932889246e-05, + "loss": 0.1292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05757030099630356, + "step": 7700, + "valid_targets_mean": 914.8, + "valid_targets_min": 485 + }, + { + "epoch": 4.247519294377067, + "grad_norm": 0.8412680108067395, + "learning_rate": 1.6067192041719065e-05, + "loss": 0.1204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06746949255466461, + "step": 7705, + "valid_targets_mean": 1047.4, + "valid_targets_min": 594 + }, + { + "epoch": 4.250275633958104, + "grad_norm": 0.7895650805453167, + "learning_rate": 1.604024218477062e-05, + "loss": 0.1379, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08395969867706299, + "step": 7710, + "valid_targets_mean": 1533.9, + "valid_targets_min": 564 + }, + { + "epoch": 4.25303197353914, + "grad_norm": 0.7818278507235246, + "learning_rate": 1.601329980896329e-05, + "loss": 0.1289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04856717213988304, + "step": 7715, + "valid_targets_mean": 983.1, + "valid_targets_min": 639 + }, + { + "epoch": 4.255788313120177, + "grad_norm": 0.7796923666324783, + "learning_rate": 1.59863649651991e-05, + "loss": 0.13, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07690586149692535, + "step": 7720, + "valid_targets_mean": 1353.0, + "valid_targets_min": 535 + }, + { + "epoch": 4.258544652701213, + "grad_norm": 0.8260899359889535, + "learning_rate": 1.5959437704365866e-05, + "loss": 0.1277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06783513724803925, + "step": 7725, + "valid_targets_mean": 1473.1, + "valid_targets_min": 1010 + }, + { + "epoch": 4.26130099228225, + "grad_norm": 0.8040776782151122, + "learning_rate": 1.5932518077337055e-05, + "loss": 0.1364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06453622877597809, + "step": 7730, + "valid_targets_mean": 1124.9, + "valid_targets_min": 592 + }, + { + "epoch": 4.2640573318632855, + "grad_norm": 0.7666361959741311, + "learning_rate": 1.590560613497172e-05, + "loss": 0.1309, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04461172968149185, + "step": 7735, + "valid_targets_mean": 955.5, + "valid_targets_min": 485 + }, + { + "epoch": 4.266813671444322, + "grad_norm": 0.8378772371796593, + "learning_rate": 1.587870192811439e-05, + "loss": 0.1322, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07552880048751831, + "step": 7740, + "valid_targets_mean": 1333.5, + "valid_targets_min": 689 + }, + { + "epoch": 4.269570011025358, + "grad_norm": 0.7891421289002315, + "learning_rate": 1.5851805507594987e-05, + "loss": 0.1396, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07169214636087418, + "step": 7745, + "valid_targets_mean": 1584.4, + "valid_targets_min": 1006 + }, + { + "epoch": 4.272326350606395, + "grad_norm": 0.8228194519126728, + "learning_rate": 1.582491692422872e-05, + "loss": 0.1444, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08759617805480957, + "step": 7750, + "valid_targets_mean": 1515.0, + "valid_targets_min": 878 + }, + { + "epoch": 4.275082690187431, + "grad_norm": 0.7842826819531248, + "learning_rate": 1.5798036228815988e-05, + "loss": 0.1334, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.062490351498126984, + "step": 7755, + "valid_targets_mean": 1360.1, + "valid_targets_min": 697 + }, + { + "epoch": 4.277839029768468, + "grad_norm": 0.7808773356188796, + "learning_rate": 1.57711634721423e-05, + "loss": 0.1257, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05287054926156998, + "step": 7760, + "valid_targets_mean": 1017.5, + "valid_targets_min": 612 + }, + { + "epoch": 4.280595369349504, + "grad_norm": 0.8952162747332363, + "learning_rate": 1.5744298704978135e-05, + "loss": 0.14, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07259555160999298, + "step": 7765, + "valid_targets_mean": 1197.0, + "valid_targets_min": 489 + }, + { + "epoch": 4.283351708930541, + "grad_norm": 0.8150425622350436, + "learning_rate": 1.5717441978078914e-05, + "loss": 0.1275, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04289630427956581, + "step": 7770, + "valid_targets_mean": 906.9, + "valid_targets_min": 568 + }, + { + "epoch": 4.286108048511577, + "grad_norm": 0.8152849301221352, + "learning_rate": 1.5690593342184854e-05, + "loss": 0.1274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059056881815195084, + "step": 7775, + "valid_targets_mean": 1167.4, + "valid_targets_min": 661 + }, + { + "epoch": 4.288864388092613, + "grad_norm": 0.7747969253344342, + "learning_rate": 1.5663752848020875e-05, + "loss": 0.1255, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06857942044734955, + "step": 7780, + "valid_targets_mean": 1421.8, + "valid_targets_min": 874 + }, + { + "epoch": 4.291620727673649, + "grad_norm": 0.7715936223240313, + "learning_rate": 1.5636920546296533e-05, + "loss": 0.134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0633678138256073, + "step": 7785, + "valid_targets_mean": 1427.1, + "valid_targets_min": 884 + }, + { + "epoch": 4.294377067254686, + "grad_norm": 0.8105427991424352, + "learning_rate": 1.561009648770589e-05, + "loss": 0.1375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06932361423969269, + "step": 7790, + "valid_targets_mean": 1310.8, + "valid_targets_min": 448 + }, + { + "epoch": 4.297133406835722, + "grad_norm": 0.8840281605854291, + "learning_rate": 1.5583280722927437e-05, + "loss": 0.1287, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06342953443527222, + "step": 7795, + "valid_targets_mean": 1178.4, + "valid_targets_min": 505 + }, + { + "epoch": 4.299889746416759, + "grad_norm": 0.8627792410611429, + "learning_rate": 1.5556473302624017e-05, + "loss": 0.1309, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06453760713338852, + "step": 7800, + "valid_targets_mean": 1342.8, + "valid_targets_min": 663 + }, + { + "epoch": 4.302646085997795, + "grad_norm": 0.8604919447611125, + "learning_rate": 1.5529674277442674e-05, + "loss": 0.1335, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061494506895542145, + "step": 7805, + "valid_targets_mean": 1205.1, + "valid_targets_min": 535 + }, + { + "epoch": 4.305402425578832, + "grad_norm": 0.80945614377353, + "learning_rate": 1.5502883698014614e-05, + "loss": 0.1356, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05609194561839104, + "step": 7810, + "valid_targets_mean": 1101.4, + "valid_targets_min": 569 + }, + { + "epoch": 4.308158765159868, + "grad_norm": 0.858091895646462, + "learning_rate": 1.547610161495508e-05, + "loss": 0.1319, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07744535803794861, + "step": 7815, + "valid_targets_mean": 1314.1, + "valid_targets_min": 518 + }, + { + "epoch": 4.3109151047409044, + "grad_norm": 0.7749808617549286, + "learning_rate": 1.5449328078863265e-05, + "loss": 0.1305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0603322833776474, + "step": 7820, + "valid_targets_mean": 1203.5, + "valid_targets_min": 818 + }, + { + "epoch": 4.31367144432194, + "grad_norm": 0.6807542979929527, + "learning_rate": 1.5422563140322208e-05, + "loss": 0.1249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06064135581254959, + "step": 7825, + "valid_targets_mean": 1295.0, + "valid_targets_min": 458 + }, + { + "epoch": 4.316427783902977, + "grad_norm": 0.8090944354595767, + "learning_rate": 1.5395806849898713e-05, + "loss": 0.1408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06897474080324173, + "step": 7830, + "valid_targets_mean": 1350.9, + "valid_targets_min": 645 + }, + { + "epoch": 4.319184123484013, + "grad_norm": 0.7760266501744029, + "learning_rate": 1.536905925814324e-05, + "loss": 0.1368, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04829941689968109, + "step": 7835, + "valid_targets_mean": 1036.2, + "valid_targets_min": 562 + }, + { + "epoch": 4.32194046306505, + "grad_norm": 0.8659541246613303, + "learning_rate": 1.5342320415589816e-05, + "loss": 0.126, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05882099270820618, + "step": 7840, + "valid_targets_mean": 1045.4, + "valid_targets_min": 613 + }, + { + "epoch": 4.324696802646086, + "grad_norm": 0.7730672294577328, + "learning_rate": 1.531559037275594e-05, + "loss": 0.1312, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05099038407206535, + "step": 7845, + "valid_targets_mean": 1178.6, + "valid_targets_min": 719 + }, + { + "epoch": 4.327453142227123, + "grad_norm": 1.014015715219083, + "learning_rate": 1.5288869180142482e-05, + "loss": 0.1403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08491168916225433, + "step": 7850, + "valid_targets_mean": 1344.9, + "valid_targets_min": 652 + }, + { + "epoch": 4.330209481808159, + "grad_norm": 0.7792705208608249, + "learning_rate": 1.5262156888233596e-05, + "loss": 0.1319, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057079702615737915, + "step": 7855, + "valid_targets_mean": 1358.0, + "valid_targets_min": 513 + }, + { + "epoch": 4.3329658213891955, + "grad_norm": 0.8256718715467936, + "learning_rate": 1.523545354749661e-05, + "loss": 0.1337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06767891347408295, + "step": 7860, + "valid_targets_mean": 1321.6, + "valid_targets_min": 638 + }, + { + "epoch": 4.335722160970231, + "grad_norm": 0.7993846126015562, + "learning_rate": 1.5208759208381952e-05, + "loss": 0.1357, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07780363410711288, + "step": 7865, + "valid_targets_mean": 1370.9, + "valid_targets_min": 693 + }, + { + "epoch": 4.338478500551268, + "grad_norm": 0.7959232151128444, + "learning_rate": 1.5182073921323044e-05, + "loss": 0.1309, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058768294751644135, + "step": 7870, + "valid_targets_mean": 999.0, + "valid_targets_min": 511 + }, + { + "epoch": 4.341234840132304, + "grad_norm": 0.8564253205997188, + "learning_rate": 1.5155397736736197e-05, + "loss": 0.1373, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06787873804569244, + "step": 7875, + "valid_targets_mean": 1327.0, + "valid_targets_min": 771 + }, + { + "epoch": 4.343991179713341, + "grad_norm": 0.8239284094188926, + "learning_rate": 1.5128730705020528e-05, + "loss": 0.1428, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0720338225364685, + "step": 7880, + "valid_targets_mean": 1376.8, + "valid_targets_min": 1115 + }, + { + "epoch": 4.346747519294377, + "grad_norm": 0.8455089920339784, + "learning_rate": 1.5102072876557867e-05, + "loss": 0.1386, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08569443225860596, + "step": 7885, + "valid_targets_mean": 1477.1, + "valid_targets_min": 888 + }, + { + "epoch": 4.349503858875414, + "grad_norm": 0.8581296767960274, + "learning_rate": 1.5075424301712645e-05, + "loss": 0.1404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06793281435966492, + "step": 7890, + "valid_targets_mean": 1301.9, + "valid_targets_min": 710 + }, + { + "epoch": 4.35226019845645, + "grad_norm": 0.7456108601140089, + "learning_rate": 1.5048785030831821e-05, + "loss": 0.1264, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06451891362667084, + "step": 7895, + "valid_targets_mean": 1291.5, + "valid_targets_min": 562 + }, + { + "epoch": 4.355016538037487, + "grad_norm": 0.7998499125498795, + "learning_rate": 1.5022155114244773e-05, + "loss": 0.132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07277597486972809, + "step": 7900, + "valid_targets_mean": 1573.1, + "valid_targets_min": 649 + }, + { + "epoch": 4.3577728776185225, + "grad_norm": 0.872295772385279, + "learning_rate": 1.499553460226321e-05, + "loss": 0.1354, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052882373332977295, + "step": 7905, + "valid_targets_mean": 982.2, + "valid_targets_min": 536 + }, + { + "epoch": 4.360529217199559, + "grad_norm": 0.7806301131694104, + "learning_rate": 1.4968923545181055e-05, + "loss": 0.1329, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06953439861536026, + "step": 7910, + "valid_targets_mean": 1252.8, + "valid_targets_min": 499 + }, + { + "epoch": 4.363285556780595, + "grad_norm": 0.7627818940006145, + "learning_rate": 1.4942321993274387e-05, + "loss": 0.1245, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056052111089229584, + "step": 7915, + "valid_targets_mean": 1108.6, + "valid_targets_min": 739 + }, + { + "epoch": 4.366041896361632, + "grad_norm": 0.894361014330129, + "learning_rate": 1.4915729996801326e-05, + "loss": 0.1423, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09272466599941254, + "step": 7920, + "valid_targets_mean": 1560.0, + "valid_targets_min": 789 + }, + { + "epoch": 4.368798235942668, + "grad_norm": 0.8302961708749704, + "learning_rate": 1.4889147606001932e-05, + "loss": 0.1402, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056999821215867996, + "step": 7925, + "valid_targets_mean": 1073.9, + "valid_targets_min": 566 + }, + { + "epoch": 4.371554575523705, + "grad_norm": 0.7608449955988178, + "learning_rate": 1.4862574871098118e-05, + "loss": 0.1348, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08092426508665085, + "step": 7930, + "valid_targets_mean": 1598.9, + "valid_targets_min": 779 + }, + { + "epoch": 4.374310915104741, + "grad_norm": 0.7275124605246203, + "learning_rate": 1.4836011842293554e-05, + "loss": 0.1317, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06903036683797836, + "step": 7935, + "valid_targets_mean": 1435.5, + "valid_targets_min": 780 + }, + { + "epoch": 4.377067254685778, + "grad_norm": 0.7576689939075518, + "learning_rate": 1.4809458569773569e-05, + "loss": 0.1166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06526580452919006, + "step": 7940, + "valid_targets_mean": 1351.5, + "valid_targets_min": 590 + }, + { + "epoch": 4.379823594266814, + "grad_norm": 1.1870565051562545, + "learning_rate": 1.4782915103705076e-05, + "loss": 0.1276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05719440430402756, + "step": 7945, + "valid_targets_mean": 1089.6, + "valid_targets_min": 477 + }, + { + "epoch": 4.38257993384785, + "grad_norm": 0.8026263335243475, + "learning_rate": 1.4756381494236443e-05, + "loss": 0.1289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07498292624950409, + "step": 7950, + "valid_targets_mean": 1345.8, + "valid_targets_min": 688 + }, + { + "epoch": 4.385336273428886, + "grad_norm": 0.8090599848164853, + "learning_rate": 1.4729857791497417e-05, + "loss": 0.1325, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07637858390808105, + "step": 7955, + "valid_targets_mean": 1753.0, + "valid_targets_min": 1002 + }, + { + "epoch": 4.388092613009923, + "grad_norm": 0.7700383515990706, + "learning_rate": 1.4703344045599037e-05, + "loss": 0.1311, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07281358540058136, + "step": 7960, + "valid_targets_mean": 1363.0, + "valid_targets_min": 517 + }, + { + "epoch": 4.390848952590959, + "grad_norm": 0.7942771316222407, + "learning_rate": 1.467684030663353e-05, + "loss": 0.1371, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057626862078905106, + "step": 7965, + "valid_targets_mean": 1133.8, + "valid_targets_min": 728 + }, + { + "epoch": 4.393605292171996, + "grad_norm": 0.8071270043684601, + "learning_rate": 1.4650346624674212e-05, + "loss": 0.1314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06773169338703156, + "step": 7970, + "valid_targets_mean": 1258.1, + "valid_targets_min": 724 + }, + { + "epoch": 4.396361631753032, + "grad_norm": 1.1548927321748312, + "learning_rate": 1.46238630497754e-05, + "loss": 0.1725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18425296247005463, + "step": 7975, + "valid_targets_mean": 5205.2, + "valid_targets_min": 1785 + }, + { + "epoch": 4.399117971334069, + "grad_norm": 0.7244009816776257, + "learning_rate": 1.4597389631972314e-05, + "loss": 0.2388, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12994727492332458, + "step": 7980, + "valid_targets_mean": 5362.8, + "valid_targets_min": 348 + }, + { + "epoch": 4.401874310915105, + "grad_norm": 0.5649488311085469, + "learning_rate": 1.4570926421280982e-05, + "loss": 0.2295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1155480146408081, + "step": 7985, + "valid_targets_mean": 4305.4, + "valid_targets_min": 1501 + }, + { + "epoch": 4.404630650496141, + "grad_norm": 0.60622649921231, + "learning_rate": 1.4544473467698165e-05, + "loss": 0.2223, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11933842301368713, + "step": 7990, + "valid_targets_mean": 3674.8, + "valid_targets_min": 993 + }, + { + "epoch": 4.407386990077177, + "grad_norm": 0.5169243570071417, + "learning_rate": 1.4518030821201223e-05, + "loss": 0.2136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10323899984359741, + "step": 7995, + "valid_targets_mean": 4216.6, + "valid_targets_min": 2254 + }, + { + "epoch": 4.410143329658214, + "grad_norm": 0.5543787041832724, + "learning_rate": 1.4491598531748053e-05, + "loss": 0.2328, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14699867367744446, + "step": 8000, + "valid_targets_mean": 5435.1, + "valid_targets_min": 1905 + }, + { + "epoch": 4.41289966923925, + "grad_norm": 0.5140868939790947, + "learning_rate": 1.4465176649276984e-05, + "loss": 0.2177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1261986494064331, + "step": 8005, + "valid_targets_mean": 6846.0, + "valid_targets_min": 3924 + }, + { + "epoch": 4.415656008820287, + "grad_norm": 0.553472045579862, + "learning_rate": 1.443876522370668e-05, + "loss": 0.2272, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12645232677459717, + "step": 8010, + "valid_targets_mean": 5263.0, + "valid_targets_min": 1538 + }, + { + "epoch": 4.418412348401323, + "grad_norm": 0.6252289877303693, + "learning_rate": 1.441236430493606e-05, + "loss": 0.2166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12275205552577972, + "step": 8015, + "valid_targets_mean": 3828.6, + "valid_targets_min": 849 + }, + { + "epoch": 4.42116868798236, + "grad_norm": 0.4830176098540896, + "learning_rate": 1.4385973942844178e-05, + "loss": 0.2308, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10831726342439651, + "step": 8020, + "valid_targets_mean": 4763.8, + "valid_targets_min": 2224 + }, + { + "epoch": 4.423925027563396, + "grad_norm": 0.6075540700718279, + "learning_rate": 1.4359594187290151e-05, + "loss": 0.2269, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1224355548620224, + "step": 8025, + "valid_targets_mean": 3492.4, + "valid_targets_min": 599 + }, + { + "epoch": 4.4266813671444325, + "grad_norm": 0.44028893420543513, + "learning_rate": 1.4333225088113058e-05, + "loss": 0.211, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11062956601381302, + "step": 8030, + "valid_targets_mean": 5308.1, + "valid_targets_min": 1593 + }, + { + "epoch": 4.429437706725468, + "grad_norm": 0.5839270323860348, + "learning_rate": 1.4306866695131837e-05, + "loss": 0.2255, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1118997260928154, + "step": 8035, + "valid_targets_mean": 3190.9, + "valid_targets_min": 1214 + }, + { + "epoch": 4.432194046306505, + "grad_norm": 0.5443924160645958, + "learning_rate": 1.4280519058145212e-05, + "loss": 0.2188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10839671641588211, + "step": 8040, + "valid_targets_mean": 3971.1, + "valid_targets_min": 817 + }, + { + "epoch": 4.434950385887541, + "grad_norm": 0.5731081741569193, + "learning_rate": 1.4254182226931574e-05, + "loss": 0.2142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12028272449970245, + "step": 8045, + "valid_targets_mean": 3689.2, + "valid_targets_min": 2305 + }, + { + "epoch": 4.437706725468578, + "grad_norm": 0.500106434718107, + "learning_rate": 1.4227856251248904e-05, + "loss": 0.2156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0955180674791336, + "step": 8050, + "valid_targets_mean": 4748.0, + "valid_targets_min": 1779 + }, + { + "epoch": 4.440463065049614, + "grad_norm": 0.6288538430911478, + "learning_rate": 1.4201541180834673e-05, + "loss": 0.2079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09927770495414734, + "step": 8055, + "valid_targets_mean": 3727.5, + "valid_targets_min": 698 + }, + { + "epoch": 4.443219404630651, + "grad_norm": 0.6088556089407335, + "learning_rate": 1.4175237065405745e-05, + "loss": 0.2176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12274210155010223, + "step": 8060, + "valid_targets_mean": 3809.9, + "valid_targets_min": 732 + }, + { + "epoch": 4.445975744211687, + "grad_norm": 0.6773652366174833, + "learning_rate": 1.4148943954658299e-05, + "loss": 0.2225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1285690814256668, + "step": 8065, + "valid_targets_mean": 3371.1, + "valid_targets_min": 790 + }, + { + "epoch": 4.448732083792724, + "grad_norm": 0.669572614129039, + "learning_rate": 1.4122661898267706e-05, + "loss": 0.2227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11980850994586945, + "step": 8070, + "valid_targets_mean": 3275.8, + "valid_targets_min": 339 + }, + { + "epoch": 4.4514884233737595, + "grad_norm": 0.6540422991408015, + "learning_rate": 1.4096390945888467e-05, + "loss": 0.215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0970693901181221, + "step": 8075, + "valid_targets_mean": 3032.0, + "valid_targets_min": 1341 + }, + { + "epoch": 4.454244762954796, + "grad_norm": 0.6022118042019735, + "learning_rate": 1.407013114715409e-05, + "loss": 0.2032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11826951801776886, + "step": 8080, + "valid_targets_mean": 3610.0, + "valid_targets_min": 1528 + }, + { + "epoch": 4.457001102535832, + "grad_norm": 0.6430812287080485, + "learning_rate": 1.4043882551677028e-05, + "loss": 0.2073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08962114155292511, + "step": 8085, + "valid_targets_mean": 3587.1, + "valid_targets_min": 719 + }, + { + "epoch": 4.459757442116869, + "grad_norm": 0.5931534069792579, + "learning_rate": 1.4017645209048554e-05, + "loss": 0.1964, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10751714557409286, + "step": 8090, + "valid_targets_mean": 3432.8, + "valid_targets_min": 432 + }, + { + "epoch": 4.462513781697905, + "grad_norm": 0.6377744777978926, + "learning_rate": 1.3991419168838683e-05, + "loss": 0.2159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11114518344402313, + "step": 8095, + "valid_targets_mean": 4320.1, + "valid_targets_min": 2227 + }, + { + "epoch": 4.465270121278942, + "grad_norm": 0.6541700759811998, + "learning_rate": 1.396520448059608e-05, + "loss": 0.2177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1063300222158432, + "step": 8100, + "valid_targets_mean": 3657.6, + "valid_targets_min": 2120 + }, + { + "epoch": 4.468026460859978, + "grad_norm": 0.6349070992100981, + "learning_rate": 1.3939001193847955e-05, + "loss": 0.2009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09412088990211487, + "step": 8105, + "valid_targets_mean": 2917.5, + "valid_targets_min": 888 + }, + { + "epoch": 4.470782800441015, + "grad_norm": 0.6062179360656665, + "learning_rate": 1.3912809358099995e-05, + "loss": 0.2059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08115215599536896, + "step": 8110, + "valid_targets_mean": 2486.2, + "valid_targets_min": 903 + }, + { + "epoch": 4.4735391400220506, + "grad_norm": 0.6721978181527393, + "learning_rate": 1.3886629022836238e-05, + "loss": 0.2081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13283686339855194, + "step": 8115, + "valid_targets_mean": 3222.9, + "valid_targets_min": 1118 + }, + { + "epoch": 4.476295479603087, + "grad_norm": 0.5720527669355672, + "learning_rate": 1.3860460237518993e-05, + "loss": 0.2039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08382523059844971, + "step": 8120, + "valid_targets_mean": 2212.4, + "valid_targets_min": 764 + }, + { + "epoch": 4.479051819184123, + "grad_norm": 0.6489394667875769, + "learning_rate": 1.3834303051588757e-05, + "loss": 0.2079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10697253048419952, + "step": 8125, + "valid_targets_mean": 3137.1, + "valid_targets_min": 1040 + }, + { + "epoch": 4.48180815876516, + "grad_norm": 0.6779467188916809, + "learning_rate": 1.3808157514464102e-05, + "loss": 0.2016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11756439507007599, + "step": 8130, + "valid_targets_mean": 3356.6, + "valid_targets_min": 1525 + }, + { + "epoch": 4.484564498346196, + "grad_norm": 0.6258810247930927, + "learning_rate": 1.3782023675541606e-05, + "loss": 0.203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12547607719898224, + "step": 8135, + "valid_targets_mean": 4259.8, + "valid_targets_min": 2386 + }, + { + "epoch": 4.487320837927233, + "grad_norm": 0.6060014007131869, + "learning_rate": 1.3755901584195728e-05, + "loss": 0.2006, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08534182608127594, + "step": 8140, + "valid_targets_mean": 2896.0, + "valid_targets_min": 921 + }, + { + "epoch": 4.490077177508269, + "grad_norm": 0.6823626896551065, + "learning_rate": 1.372979128977875e-05, + "loss": 0.1999, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11366604268550873, + "step": 8145, + "valid_targets_mean": 3591.9, + "valid_targets_min": 285 + }, + { + "epoch": 4.492833517089306, + "grad_norm": 0.6125003087336305, + "learning_rate": 1.3703692841620653e-05, + "loss": 0.2007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10670024156570435, + "step": 8150, + "valid_targets_mean": 3798.2, + "valid_targets_min": 2294 + }, + { + "epoch": 4.495589856670342, + "grad_norm": 0.6444042157595978, + "learning_rate": 1.367760628902904e-05, + "loss": 0.2146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09666654467582703, + "step": 8155, + "valid_targets_mean": 2942.4, + "valid_targets_min": 1344 + }, + { + "epoch": 4.498346196251378, + "grad_norm": 0.6236202620103812, + "learning_rate": 1.3651531681289048e-05, + "loss": 0.2028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09723399579524994, + "step": 8160, + "valid_targets_mean": 3422.4, + "valid_targets_min": 1585 + }, + { + "epoch": 4.501102535832414, + "grad_norm": 0.5628999447978364, + "learning_rate": 1.3625469067663237e-05, + "loss": 0.2018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10036353766918182, + "step": 8165, + "valid_targets_mean": 4884.0, + "valid_targets_min": 2276 + }, + { + "epoch": 4.503858875413451, + "grad_norm": 0.6576329246872736, + "learning_rate": 1.3599418497391512e-05, + "loss": 0.2049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11736799776554108, + "step": 8170, + "valid_targets_mean": 3428.4, + "valid_targets_min": 1488 + }, + { + "epoch": 4.506615214994487, + "grad_norm": 0.7176058011863913, + "learning_rate": 1.3573380019691019e-05, + "loss": 0.2079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10288487374782562, + "step": 8175, + "valid_targets_mean": 4393.0, + "valid_targets_min": 1557 + }, + { + "epoch": 4.509371554575524, + "grad_norm": 0.6265123159461893, + "learning_rate": 1.3547353683756056e-05, + "loss": 0.2028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1059994250535965, + "step": 8180, + "valid_targets_mean": 3035.8, + "valid_targets_min": 1306 + }, + { + "epoch": 4.51212789415656, + "grad_norm": 0.583008446578367, + "learning_rate": 1.3521339538758e-05, + "loss": 0.2021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09527242183685303, + "step": 8185, + "valid_targets_mean": 4112.4, + "valid_targets_min": 1897 + }, + { + "epoch": 4.514884233737597, + "grad_norm": 0.622362406457088, + "learning_rate": 1.3495337633845175e-05, + "loss": 0.2038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09913437068462372, + "step": 8190, + "valid_targets_mean": 3556.9, + "valid_targets_min": 2236 + }, + { + "epoch": 4.517640573318633, + "grad_norm": 0.6447818681200578, + "learning_rate": 1.3469348018142786e-05, + "loss": 0.1971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10169985145330429, + "step": 8195, + "valid_targets_mean": 3293.0, + "valid_targets_min": 1117 + }, + { + "epoch": 4.5203969128996695, + "grad_norm": 0.676787494294692, + "learning_rate": 1.3443370740752823e-05, + "loss": 0.1986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08845679461956024, + "step": 8200, + "valid_targets_mean": 2474.5, + "valid_targets_min": 680 + }, + { + "epoch": 4.523153252480705, + "grad_norm": 0.5562708567038226, + "learning_rate": 1.3417405850753968e-05, + "loss": 0.208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09789156913757324, + "step": 8205, + "valid_targets_mean": 4263.0, + "valid_targets_min": 1859 + }, + { + "epoch": 4.525909592061742, + "grad_norm": 0.6050515854113211, + "learning_rate": 1.3391453397201493e-05, + "loss": 0.204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11567363888025284, + "step": 8210, + "valid_targets_mean": 3843.9, + "valid_targets_min": 1646 + }, + { + "epoch": 4.528665931642778, + "grad_norm": 0.6717022117091497, + "learning_rate": 1.3365513429127172e-05, + "loss": 0.2183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09953713417053223, + "step": 8215, + "valid_targets_mean": 2990.8, + "valid_targets_min": 773 + }, + { + "epoch": 4.531422271223815, + "grad_norm": 0.6541310884511502, + "learning_rate": 1.33395859955392e-05, + "loss": 0.2001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10146254301071167, + "step": 8220, + "valid_targets_mean": 2829.9, + "valid_targets_min": 823 + }, + { + "epoch": 4.534178610804851, + "grad_norm": 0.5971394267571762, + "learning_rate": 1.3313671145422077e-05, + "loss": 0.2001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10028070211410522, + "step": 8225, + "valid_targets_mean": 3993.1, + "valid_targets_min": 1793 + }, + { + "epoch": 4.536934950385888, + "grad_norm": 0.9370097278314133, + "learning_rate": 1.328776892773655e-05, + "loss": 0.2057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10061026364564896, + "step": 8230, + "valid_targets_mean": 2828.6, + "valid_targets_min": 843 + }, + { + "epoch": 4.539691289966924, + "grad_norm": 0.5855432175182074, + "learning_rate": 1.3261879391419478e-05, + "loss": 0.2019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0858074277639389, + "step": 8235, + "valid_targets_mean": 3449.0, + "valid_targets_min": 1971 + }, + { + "epoch": 4.5424476295479606, + "grad_norm": 0.6262917340071907, + "learning_rate": 1.3236002585383774e-05, + "loss": 0.2001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10498316586017609, + "step": 8240, + "valid_targets_mean": 4142.8, + "valid_targets_min": 1022 + }, + { + "epoch": 4.5452039691289965, + "grad_norm": 0.6062428132890755, + "learning_rate": 1.3210138558518295e-05, + "loss": 0.1923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09517504274845123, + "step": 8245, + "valid_targets_mean": 2985.8, + "valid_targets_min": 1165 + }, + { + "epoch": 4.547960308710033, + "grad_norm": 0.5879648077160533, + "learning_rate": 1.3184287359687751e-05, + "loss": 0.1856, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.090573750436306, + "step": 8250, + "valid_targets_mean": 3336.4, + "valid_targets_min": 959 + }, + { + "epoch": 4.550716648291069, + "grad_norm": 0.6682786391097779, + "learning_rate": 1.3158449037732628e-05, + "loss": 0.2213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0930638462305069, + "step": 8255, + "valid_targets_mean": 2212.0, + "valid_targets_min": 1197 + }, + { + "epoch": 4.553472987872106, + "grad_norm": 0.5169715969916107, + "learning_rate": 1.3132623641469074e-05, + "loss": 0.1965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0858829915523529, + "step": 8260, + "valid_targets_mean": 5110.5, + "valid_targets_min": 2482 + }, + { + "epoch": 4.556229327453142, + "grad_norm": 0.614308338511065, + "learning_rate": 1.3106811219688815e-05, + "loss": 0.2129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08905627578496933, + "step": 8265, + "valid_targets_mean": 2911.6, + "valid_targets_min": 1064 + }, + { + "epoch": 4.558985667034179, + "grad_norm": 0.6529544873875166, + "learning_rate": 1.3081011821159075e-05, + "loss": 0.2031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13064846396446228, + "step": 8270, + "valid_targets_mean": 4130.2, + "valid_targets_min": 1181 + }, + { + "epoch": 4.561742006615215, + "grad_norm": 0.6255561833268264, + "learning_rate": 1.3055225494622459e-05, + "loss": 0.2099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10430508852005005, + "step": 8275, + "valid_targets_mean": 3570.4, + "valid_targets_min": 1329 + }, + { + "epoch": 4.564498346196252, + "grad_norm": 0.6075532049073561, + "learning_rate": 1.3029452288796886e-05, + "loss": 0.2021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11155767738819122, + "step": 8280, + "valid_targets_mean": 3641.4, + "valid_targets_min": 1818 + }, + { + "epoch": 4.5672546857772875, + "grad_norm": 0.6454714320084324, + "learning_rate": 1.3003692252375486e-05, + "loss": 0.1938, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11650878190994263, + "step": 8285, + "valid_targets_mean": 3479.8, + "valid_targets_min": 2410 + }, + { + "epoch": 4.570011025358324, + "grad_norm": 0.6484242000852743, + "learning_rate": 1.2977945434026504e-05, + "loss": 0.1934, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10387968271970749, + "step": 8290, + "valid_targets_mean": 3051.8, + "valid_targets_min": 1935 + }, + { + "epoch": 4.57276736493936, + "grad_norm": 0.6090892896044147, + "learning_rate": 1.2952211882393211e-05, + "loss": 0.1897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09731649607419968, + "step": 8295, + "valid_targets_mean": 3740.4, + "valid_targets_min": 1836 + }, + { + "epoch": 4.575523704520397, + "grad_norm": 0.6222681043741118, + "learning_rate": 1.292649164609381e-05, + "loss": 0.1866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08072851598262787, + "step": 8300, + "valid_targets_mean": 2487.6, + "valid_targets_min": 819 + }, + { + "epoch": 4.578280044101433, + "grad_norm": 0.726141908928948, + "learning_rate": 1.2900784773721365e-05, + "loss": 0.1943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11705812066793442, + "step": 8305, + "valid_targets_mean": 3737.6, + "valid_targets_min": 1039 + }, + { + "epoch": 4.58103638368247, + "grad_norm": 0.6438713981037394, + "learning_rate": 1.2875091313843674e-05, + "loss": 0.1809, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09208910167217255, + "step": 8310, + "valid_targets_mean": 3166.1, + "valid_targets_min": 756 + }, + { + "epoch": 4.583792723263506, + "grad_norm": 0.6856641686057695, + "learning_rate": 1.2849411315003199e-05, + "loss": 0.1913, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11032165586948395, + "step": 8315, + "valid_targets_mean": 3918.5, + "valid_targets_min": 577 + }, + { + "epoch": 4.586549062844543, + "grad_norm": 0.5951366839672051, + "learning_rate": 1.2823744825716974e-05, + "loss": 0.2208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08413458615541458, + "step": 8320, + "valid_targets_mean": 3691.4, + "valid_targets_min": 1861 + }, + { + "epoch": 4.589305402425579, + "grad_norm": 0.6559564199255264, + "learning_rate": 1.2798091894476503e-05, + "loss": 0.1921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11756572872400284, + "step": 8325, + "valid_targets_mean": 3579.1, + "valid_targets_min": 1118 + }, + { + "epoch": 4.592061742006615, + "grad_norm": 0.8232209157633267, + "learning_rate": 1.2772452569747685e-05, + "loss": 0.2001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10102914273738861, + "step": 8330, + "valid_targets_mean": 4181.9, + "valid_targets_min": 2142 + }, + { + "epoch": 4.594818081587651, + "grad_norm": 0.5809827575204806, + "learning_rate": 1.2746826899970706e-05, + "loss": 0.19, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08371060341596603, + "step": 8335, + "valid_targets_mean": 3453.2, + "valid_targets_min": 1487 + }, + { + "epoch": 4.597574421168688, + "grad_norm": 0.7691166619205296, + "learning_rate": 1.2721214933559947e-05, + "loss": 0.1965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10350465029478073, + "step": 8340, + "valid_targets_mean": 3903.1, + "valid_targets_min": 1381 + }, + { + "epoch": 4.600330760749724, + "grad_norm": 0.6922565471601584, + "learning_rate": 1.269561671890391e-05, + "loss": 0.1976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11537542939186096, + "step": 8345, + "valid_targets_mean": 3452.1, + "valid_targets_min": 1359 + }, + { + "epoch": 4.603087100330761, + "grad_norm": 0.5703825843213065, + "learning_rate": 1.2670032304365116e-05, + "loss": 0.1847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09146757423877716, + "step": 8350, + "valid_targets_mean": 3630.0, + "valid_targets_min": 917 + }, + { + "epoch": 4.605843439911797, + "grad_norm": 0.6060810347756957, + "learning_rate": 1.264446173828001e-05, + "loss": 0.2014, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08412057906389236, + "step": 8355, + "valid_targets_mean": 3292.2, + "valid_targets_min": 1554 + }, + { + "epoch": 4.608599779492834, + "grad_norm": 0.6153812123215717, + "learning_rate": 1.2618905068958867e-05, + "loss": 0.1917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10408687591552734, + "step": 8360, + "valid_targets_mean": 3273.4, + "valid_targets_min": 1575 + }, + { + "epoch": 4.61135611907387, + "grad_norm": 0.6895793617789204, + "learning_rate": 1.2593362344685717e-05, + "loss": 0.1936, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11807402223348618, + "step": 8365, + "valid_targets_mean": 3361.5, + "valid_targets_min": 1838 + }, + { + "epoch": 4.6141124586549065, + "grad_norm": 0.6748071334707851, + "learning_rate": 1.2567833613718237e-05, + "loss": 0.2073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10716613382101059, + "step": 8370, + "valid_targets_mean": 3372.5, + "valid_targets_min": 1829 + }, + { + "epoch": 4.616868798235942, + "grad_norm": 0.6602075097787454, + "learning_rate": 1.2542318924287678e-05, + "loss": 0.1988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11416689306497574, + "step": 8375, + "valid_targets_mean": 3712.1, + "valid_targets_min": 1374 + }, + { + "epoch": 4.619625137816979, + "grad_norm": 0.7066012051848862, + "learning_rate": 1.2516818324598742e-05, + "loss": 0.2077, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1081610918045044, + "step": 8380, + "valid_targets_mean": 3077.8, + "valid_targets_min": 1295 + }, + { + "epoch": 4.622381477398015, + "grad_norm": 0.5700105974148011, + "learning_rate": 1.2491331862829532e-05, + "loss": 0.1976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07880906760692596, + "step": 8385, + "valid_targets_mean": 3564.4, + "valid_targets_min": 1563 + }, + { + "epoch": 4.625137816979052, + "grad_norm": 0.6157500355705191, + "learning_rate": 1.2465859587131428e-05, + "loss": 0.2146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09651237726211548, + "step": 8390, + "valid_targets_mean": 3693.2, + "valid_targets_min": 2210 + }, + { + "epoch": 4.627894156560088, + "grad_norm": 0.6468003661336467, + "learning_rate": 1.2440401545629009e-05, + "loss": 0.1984, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08608653396368027, + "step": 8395, + "valid_targets_mean": 3202.2, + "valid_targets_min": 1413 + }, + { + "epoch": 4.630650496141125, + "grad_norm": 0.6298958473064129, + "learning_rate": 1.2414957786419967e-05, + "loss": 0.2032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09532776474952698, + "step": 8400, + "valid_targets_mean": 3205.2, + "valid_targets_min": 2248 + }, + { + "epoch": 4.633406835722161, + "grad_norm": 0.6904559317105118, + "learning_rate": 1.2389528357575013e-05, + "loss": 0.2018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10766679793596268, + "step": 8405, + "valid_targets_mean": 3377.1, + "valid_targets_min": 1704 + }, + { + "epoch": 4.6361631753031975, + "grad_norm": 0.6455809735552381, + "learning_rate": 1.2364113307137773e-05, + "loss": 0.1947, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08069120347499847, + "step": 8410, + "valid_targets_mean": 3006.8, + "valid_targets_min": 361 + }, + { + "epoch": 4.6389195148842335, + "grad_norm": 0.6531295799208233, + "learning_rate": 1.2338712683124717e-05, + "loss": 0.1995, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08312135934829712, + "step": 8415, + "valid_targets_mean": 3035.9, + "valid_targets_min": 1200 + }, + { + "epoch": 4.64167585446527, + "grad_norm": 0.6696114498501199, + "learning_rate": 1.231332653352505e-05, + "loss": 0.2005, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11011239886283875, + "step": 8420, + "valid_targets_mean": 3808.2, + "valid_targets_min": 1015 + }, + { + "epoch": 4.644432194046306, + "grad_norm": 0.6607290712419157, + "learning_rate": 1.2287954906300638e-05, + "loss": 0.192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09807047992944717, + "step": 8425, + "valid_targets_mean": 2776.1, + "valid_targets_min": 1355 + }, + { + "epoch": 4.647188533627343, + "grad_norm": 0.6513704541328652, + "learning_rate": 1.2262597849385908e-05, + "loss": 0.1849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08000001311302185, + "step": 8430, + "valid_targets_mean": 4237.1, + "valid_targets_min": 959 + }, + { + "epoch": 4.649944873208379, + "grad_norm": 0.679368098130922, + "learning_rate": 1.2237255410687766e-05, + "loss": 0.1987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08941663801670074, + "step": 8435, + "valid_targets_mean": 3085.6, + "valid_targets_min": 970 + }, + { + "epoch": 4.652701212789416, + "grad_norm": 0.6166954002580433, + "learning_rate": 1.2211927638085487e-05, + "loss": 0.1837, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07669895887374878, + "step": 8440, + "valid_targets_mean": 2649.6, + "valid_targets_min": 515 + }, + { + "epoch": 4.655457552370452, + "grad_norm": 0.6366937106166042, + "learning_rate": 1.218661457943065e-05, + "loss": 0.1992, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10280340164899826, + "step": 8445, + "valid_targets_mean": 3388.6, + "valid_targets_min": 1339 + }, + { + "epoch": 4.658213891951489, + "grad_norm": 0.7163815559940976, + "learning_rate": 1.2161316282547024e-05, + "loss": 0.1952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09859108924865723, + "step": 8450, + "valid_targets_mean": 2546.1, + "valid_targets_min": 1250 + }, + { + "epoch": 4.6609702315325245, + "grad_norm": 0.6020840919046166, + "learning_rate": 1.2136032795230492e-05, + "loss": 0.1974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11548559367656708, + "step": 8455, + "valid_targets_mean": 5238.6, + "valid_targets_min": 2699 + }, + { + "epoch": 4.663726571113561, + "grad_norm": 0.7592103988908578, + "learning_rate": 1.211076416524897e-05, + "loss": 0.1898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11944218724966049, + "step": 8460, + "valid_targets_mean": 2922.8, + "valid_targets_min": 1315 + }, + { + "epoch": 4.666482910694597, + "grad_norm": 0.6135787171932134, + "learning_rate": 1.2085510440342282e-05, + "loss": 0.2008, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09297943115234375, + "step": 8465, + "valid_targets_mean": 3901.6, + "valid_targets_min": 1929 + }, + { + "epoch": 4.669239250275634, + "grad_norm": 0.5604243741891163, + "learning_rate": 1.2060271668222115e-05, + "loss": 0.1923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08999551087617874, + "step": 8470, + "valid_targets_mean": 5600.1, + "valid_targets_min": 734 + }, + { + "epoch": 4.67199558985667, + "grad_norm": 0.5595304794756107, + "learning_rate": 1.2035047896571884e-05, + "loss": 0.1876, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0923205316066742, + "step": 8475, + "valid_targets_mean": 5048.9, + "valid_targets_min": 1197 + }, + { + "epoch": 4.674751929437707, + "grad_norm": 0.5934135688693799, + "learning_rate": 1.2009839173046673e-05, + "loss": 0.1877, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09877340495586395, + "step": 8480, + "valid_targets_mean": 3522.4, + "valid_targets_min": 590 + }, + { + "epoch": 4.677508269018743, + "grad_norm": 0.5983634504302565, + "learning_rate": 1.1984645545273145e-05, + "loss": 0.1917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09932804107666016, + "step": 8485, + "valid_targets_mean": 3415.2, + "valid_targets_min": 760 + }, + { + "epoch": 4.68026460859978, + "grad_norm": 0.6205393238606864, + "learning_rate": 1.1959467060849432e-05, + "loss": 0.1869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08592885732650757, + "step": 8490, + "valid_targets_mean": 3609.6, + "valid_targets_min": 541 + }, + { + "epoch": 4.683020948180816, + "grad_norm": 0.6874401924553921, + "learning_rate": 1.1934303767345057e-05, + "loss": 0.1819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09382587671279907, + "step": 8495, + "valid_targets_mean": 2825.2, + "valid_targets_min": 1437 + }, + { + "epoch": 4.685777287761852, + "grad_norm": 0.6497717660852654, + "learning_rate": 1.1909155712300844e-05, + "loss": 0.1844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11063970625400543, + "step": 8500, + "valid_targets_mean": 4688.1, + "valid_targets_min": 3307 + }, + { + "epoch": 4.688533627342888, + "grad_norm": 0.7096191606016146, + "learning_rate": 1.1884022943228824e-05, + "loss": 0.1821, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10157299786806107, + "step": 8505, + "valid_targets_mean": 3752.9, + "valid_targets_min": 737 + }, + { + "epoch": 4.691289966923925, + "grad_norm": 0.7643985064800509, + "learning_rate": 1.1858905507612156e-05, + "loss": 0.2011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10488620400428772, + "step": 8510, + "valid_targets_mean": 2480.1, + "valid_targets_min": 1098 + }, + { + "epoch": 4.694046306504961, + "grad_norm": 0.656298575560868, + "learning_rate": 1.1833803452905025e-05, + "loss": 0.1977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09489640593528748, + "step": 8515, + "valid_targets_mean": 2641.9, + "valid_targets_min": 1101 + }, + { + "epoch": 4.696802646085998, + "grad_norm": 0.6813304033081931, + "learning_rate": 1.1808716826532556e-05, + "loss": 0.183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10239865630865097, + "step": 8520, + "valid_targets_mean": 3319.1, + "valid_targets_min": 1842 + }, + { + "epoch": 4.699558985667034, + "grad_norm": 0.6161572057827146, + "learning_rate": 1.1783645675890725e-05, + "loss": 0.18, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07892011106014252, + "step": 8525, + "valid_targets_mean": 2775.6, + "valid_targets_min": 476 + }, + { + "epoch": 4.702315325248071, + "grad_norm": 0.7297733413230764, + "learning_rate": 1.175859004834627e-05, + "loss": 0.1947, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09297402948141098, + "step": 8530, + "valid_targets_mean": 2842.9, + "valid_targets_min": 516 + }, + { + "epoch": 4.705071664829107, + "grad_norm": 0.6943979038071617, + "learning_rate": 1.1733549991236608e-05, + "loss": 0.1888, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10295884311199188, + "step": 8535, + "valid_targets_mean": 3877.9, + "valid_targets_min": 1074 + }, + { + "epoch": 4.7078280044101435, + "grad_norm": 0.7105045078946002, + "learning_rate": 1.1708525551869729e-05, + "loss": 0.2004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09197092056274414, + "step": 8540, + "valid_targets_mean": 2300.0, + "valid_targets_min": 235 + }, + { + "epoch": 4.710584343991179, + "grad_norm": 0.6286944748853168, + "learning_rate": 1.1683516777524127e-05, + "loss": 0.1931, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08211054652929306, + "step": 8545, + "valid_targets_mean": 2961.6, + "valid_targets_min": 1851 + }, + { + "epoch": 4.713340683572216, + "grad_norm": 0.6061387615793724, + "learning_rate": 1.1658523715448683e-05, + "loss": 0.1877, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09564010798931122, + "step": 8550, + "valid_targets_mean": 3412.0, + "valid_targets_min": 1698 + }, + { + "epoch": 4.716097023153252, + "grad_norm": 0.6287864195282215, + "learning_rate": 1.1633546412862612e-05, + "loss": 0.1995, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08014830946922302, + "step": 8555, + "valid_targets_mean": 2786.1, + "valid_targets_min": 983 + }, + { + "epoch": 4.718853362734289, + "grad_norm": 0.6602189735509278, + "learning_rate": 1.1608584916955348e-05, + "loss": 0.1818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07288560271263123, + "step": 8560, + "valid_targets_mean": 1835.6, + "valid_targets_min": 631 + }, + { + "epoch": 4.721609702315325, + "grad_norm": 0.8583629416783974, + "learning_rate": 1.1583639274886461e-05, + "loss": 0.1009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0320083424448967, + "step": 8565, + "valid_targets_mean": 2832.5, + "valid_targets_min": 829 + }, + { + "epoch": 4.724366041896362, + "grad_norm": 0.379112365579214, + "learning_rate": 1.155870953378556e-05, + "loss": 0.0802, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03570215031504631, + "step": 8570, + "valid_targets_mean": 3424.2, + "valid_targets_min": 2780 + }, + { + "epoch": 4.727122381477398, + "grad_norm": 0.3888345044302394, + "learning_rate": 1.1533795740752228e-05, + "loss": 0.0779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04234263300895691, + "step": 8575, + "valid_targets_mean": 3527.0, + "valid_targets_min": 2121 + }, + { + "epoch": 4.7298787210584345, + "grad_norm": 0.39948799908753885, + "learning_rate": 1.1508897942855902e-05, + "loss": 0.0921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03544847294688225, + "step": 8580, + "valid_targets_mean": 2447.4, + "valid_targets_min": 523 + }, + { + "epoch": 4.7326350606394705, + "grad_norm": 0.4178120625821207, + "learning_rate": 1.1484016187135815e-05, + "loss": 0.077, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04247087985277176, + "step": 8585, + "valid_targets_mean": 2761.6, + "valid_targets_min": 885 + }, + { + "epoch": 4.735391400220507, + "grad_norm": 0.9881328587674453, + "learning_rate": 1.1459150520600873e-05, + "loss": 0.1038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0798206478357315, + "step": 8590, + "valid_targets_mean": 1135.5, + "valid_targets_min": 556 + }, + { + "epoch": 4.738147739801543, + "grad_norm": 0.4469826542081459, + "learning_rate": 1.143430099022961e-05, + "loss": 0.1165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031181320548057556, + "step": 8595, + "valid_targets_mean": 2661.6, + "valid_targets_min": 643 + }, + { + "epoch": 4.74090407938258, + "grad_norm": 0.4276023922882361, + "learning_rate": 1.1409467642970053e-05, + "loss": 0.0865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039612822234630585, + "step": 8600, + "valid_targets_mean": 2996.8, + "valid_targets_min": 942 + }, + { + "epoch": 4.743660418963616, + "grad_norm": 0.5468821480369525, + "learning_rate": 1.1384650525739662e-05, + "loss": 0.1099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09219682216644287, + "step": 8605, + "valid_targets_mean": 1892.5, + "valid_targets_min": 807 + }, + { + "epoch": 4.746416758544653, + "grad_norm": 0.3141756010575541, + "learning_rate": 1.1359849685425232e-05, + "loss": 0.0804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04366514831781387, + "step": 8610, + "valid_targets_mean": 5422.9, + "valid_targets_min": 633 + }, + { + "epoch": 4.749173098125689, + "grad_norm": 0.4005847654258223, + "learning_rate": 1.1335065168882807e-05, + "loss": 0.0768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04191962629556656, + "step": 8615, + "valid_targets_mean": 2580.2, + "valid_targets_min": 821 + }, + { + "epoch": 4.751929437706726, + "grad_norm": 0.2956808571993688, + "learning_rate": 1.1310297022937589e-05, + "loss": 0.0792, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03610297292470932, + "step": 8620, + "valid_targets_mean": 3730.6, + "valid_targets_min": 1091 + }, + { + "epoch": 4.7546857772877615, + "grad_norm": 0.4450828101955474, + "learning_rate": 1.1285545294383855e-05, + "loss": 0.0941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04631480574607849, + "step": 8625, + "valid_targets_mean": 2870.5, + "valid_targets_min": 684 + }, + { + "epoch": 4.757442116868798, + "grad_norm": 0.34869706624600944, + "learning_rate": 1.1260810029984861e-05, + "loss": 0.1001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03238993138074875, + "step": 8630, + "valid_targets_mean": 3119.1, + "valid_targets_min": 550 + }, + { + "epoch": 4.760198456449834, + "grad_norm": 0.4314439934338067, + "learning_rate": 1.1236091276472751e-05, + "loss": 0.0802, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04197119176387787, + "step": 8635, + "valid_targets_mean": 2393.1, + "valid_targets_min": 563 + }, + { + "epoch": 4.762954796030871, + "grad_norm": 0.3032746319132246, + "learning_rate": 1.1211389080548499e-05, + "loss": 0.0704, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03209776058793068, + "step": 8640, + "valid_targets_mean": 4064.1, + "valid_targets_min": 3326 + }, + { + "epoch": 4.765711135611907, + "grad_norm": 0.4554303790541253, + "learning_rate": 1.1186703488881773e-05, + "loss": 0.0755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05604122579097748, + "step": 8645, + "valid_targets_mean": 1677.4, + "valid_targets_min": 619 + }, + { + "epoch": 4.768467475192944, + "grad_norm": 0.5616610732348284, + "learning_rate": 1.116203454811088e-05, + "loss": 0.0908, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05296391248703003, + "step": 8650, + "valid_targets_mean": 2418.4, + "valid_targets_min": 692 + }, + { + "epoch": 4.77122381477398, + "grad_norm": 0.6225502035359628, + "learning_rate": 1.1137382304842665e-05, + "loss": 0.0867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055542491376399994, + "step": 8655, + "valid_targets_mean": 3002.8, + "valid_targets_min": 1914 + }, + { + "epoch": 4.773980154355017, + "grad_norm": 0.46311061886455773, + "learning_rate": 1.1112746805652432e-05, + "loss": 0.0872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029007859528064728, + "step": 8660, + "valid_targets_mean": 2430.4, + "valid_targets_min": 797 + }, + { + "epoch": 4.776736493936053, + "grad_norm": 0.5154165460144824, + "learning_rate": 1.1088128097083841e-05, + "loss": 0.0945, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05685199052095413, + "step": 8665, + "valid_targets_mean": 2037.0, + "valid_targets_min": 952 + }, + { + "epoch": 4.779492833517089, + "grad_norm": 0.3721866818954764, + "learning_rate": 1.106352622564884e-05, + "loss": 0.1065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04230587184429169, + "step": 8670, + "valid_targets_mean": 2504.5, + "valid_targets_min": 638 + }, + { + "epoch": 4.782249173098125, + "grad_norm": 0.38055810541807206, + "learning_rate": 1.1038941237827565e-05, + "loss": 0.076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04477271810173988, + "step": 8675, + "valid_targets_mean": 2675.4, + "valid_targets_min": 1251 + }, + { + "epoch": 4.785005512679162, + "grad_norm": 0.5654613885750333, + "learning_rate": 1.1014373180068246e-05, + "loss": 0.1164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08524647355079651, + "step": 8680, + "valid_targets_mean": 2130.9, + "valid_targets_min": 696 + }, + { + "epoch": 4.787761852260198, + "grad_norm": 0.4585497466187221, + "learning_rate": 1.0989822098787126e-05, + "loss": 0.0972, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05585373938083649, + "step": 8685, + "valid_targets_mean": 2731.6, + "valid_targets_min": 607 + }, + { + "epoch": 4.790518191841235, + "grad_norm": 0.4787907659798296, + "learning_rate": 1.09652880403684e-05, + "loss": 0.0809, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055716097354888916, + "step": 8690, + "valid_targets_mean": 2683.6, + "valid_targets_min": 888 + }, + { + "epoch": 4.793274531422271, + "grad_norm": 0.6446424865642185, + "learning_rate": 1.0940771051164072e-05, + "loss": 0.2349, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2567692697048187, + "step": 8695, + "valid_targets_mean": 2824.9, + "valid_targets_min": 1166 + }, + { + "epoch": 4.796030871003308, + "grad_norm": 0.47012102501740716, + "learning_rate": 1.0916271177493907e-05, + "loss": 0.1189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03943303972482681, + "step": 8700, + "valid_targets_mean": 2085.8, + "valid_targets_min": 716 + }, + { + "epoch": 4.798787210584344, + "grad_norm": 0.39232383575144936, + "learning_rate": 1.0891788465645338e-05, + "loss": 0.0964, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052037909626960754, + "step": 8705, + "valid_targets_mean": 3255.1, + "valid_targets_min": 612 + }, + { + "epoch": 4.8015435501653805, + "grad_norm": 0.5101399621473813, + "learning_rate": 1.0867322961873373e-05, + "loss": 0.1726, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07829974591732025, + "step": 8710, + "valid_targets_mean": 2770.4, + "valid_targets_min": 681 + }, + { + "epoch": 4.804299889746416, + "grad_norm": 0.4272610141708178, + "learning_rate": 1.0842874712400499e-05, + "loss": 0.0829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030818849802017212, + "step": 8715, + "valid_targets_mean": 2269.5, + "valid_targets_min": 825 + }, + { + "epoch": 4.807056229327453, + "grad_norm": 0.3929775052732992, + "learning_rate": 1.0818443763416622e-05, + "loss": 0.0815, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04551367834210396, + "step": 8720, + "valid_targets_mean": 5000.9, + "valid_targets_min": 2660 + }, + { + "epoch": 4.809812568908489, + "grad_norm": 0.3389030541755133, + "learning_rate": 1.0794030161078948e-05, + "loss": 0.0874, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03145987540483475, + "step": 8725, + "valid_targets_mean": 3396.2, + "valid_targets_min": 991 + }, + { + "epoch": 4.812568908489526, + "grad_norm": 0.42658861549798743, + "learning_rate": 1.0769633951511915e-05, + "loss": 0.0783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0506431981921196, + "step": 8730, + "valid_targets_mean": 3447.6, + "valid_targets_min": 868 + }, + { + "epoch": 4.815325248070562, + "grad_norm": 0.4261107958193727, + "learning_rate": 1.0745255180807094e-05, + "loss": 0.0793, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04259622097015381, + "step": 8735, + "valid_targets_mean": 3692.9, + "valid_targets_min": 668 + }, + { + "epoch": 4.818081587651599, + "grad_norm": 0.3532008459574235, + "learning_rate": 1.0720893895023134e-05, + "loss": 0.0713, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031122852116823196, + "step": 8740, + "valid_targets_mean": 3203.4, + "valid_targets_min": 1761 + }, + { + "epoch": 4.820837927232635, + "grad_norm": 0.38030015569872533, + "learning_rate": 1.069655014018562e-05, + "loss": 0.0798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036430153995752335, + "step": 8745, + "valid_targets_mean": 3626.9, + "valid_targets_min": 2530 + }, + { + "epoch": 4.8235942668136715, + "grad_norm": 0.45364450761180325, + "learning_rate": 1.067222396228703e-05, + "loss": 0.0697, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05071263387799263, + "step": 8750, + "valid_targets_mean": 3387.9, + "valid_targets_min": 1233 + }, + { + "epoch": 4.8263506063947075, + "grad_norm": 0.4649763035674763, + "learning_rate": 1.0647915407286632e-05, + "loss": 0.1141, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02329001948237419, + "step": 8755, + "valid_targets_mean": 3254.9, + "valid_targets_min": 619 + }, + { + "epoch": 4.829106945975744, + "grad_norm": 0.38612268821681583, + "learning_rate": 1.0623624521110397e-05, + "loss": 0.0652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033282119780778885, + "step": 8760, + "valid_targets_mean": 3229.9, + "valid_targets_min": 585 + }, + { + "epoch": 4.83186328555678, + "grad_norm": 0.42426248197885075, + "learning_rate": 1.059935134965092e-05, + "loss": 0.0714, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038710035383701324, + "step": 8765, + "valid_targets_mean": 3873.1, + "valid_targets_min": 3327 + }, + { + "epoch": 4.834619625137817, + "grad_norm": 0.43017288253573877, + "learning_rate": 1.057509593876732e-05, + "loss": 0.0703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.021611008793115616, + "step": 8770, + "valid_targets_mean": 1698.1, + "valid_targets_min": 523 + }, + { + "epoch": 4.837375964718853, + "grad_norm": 0.6751012025526197, + "learning_rate": 1.0550858334285165e-05, + "loss": 0.1091, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08502981811761856, + "step": 8775, + "valid_targets_mean": 1557.5, + "valid_targets_min": 194 + }, + { + "epoch": 4.84013230429989, + "grad_norm": 0.36015246299869036, + "learning_rate": 1.0526638581996378e-05, + "loss": 0.0753, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0459870807826519, + "step": 8780, + "valid_targets_mean": 3791.4, + "valid_targets_min": 3033 + }, + { + "epoch": 4.842888643880926, + "grad_norm": 0.49928722425706223, + "learning_rate": 1.0502436727659169e-05, + "loss": 0.1139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05377185344696045, + "step": 8785, + "valid_targets_mean": 2261.0, + "valid_targets_min": 959 + }, + { + "epoch": 4.845644983461963, + "grad_norm": 0.41981611998693485, + "learning_rate": 1.0478252816997918e-05, + "loss": 0.0808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039104267954826355, + "step": 8790, + "valid_targets_mean": 3329.0, + "valid_targets_min": 1025 + }, + { + "epoch": 4.8484013230429985, + "grad_norm": 0.6120090940577467, + "learning_rate": 1.0454086895703108e-05, + "loss": 0.1065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06383004039525986, + "step": 8795, + "valid_targets_mean": 1613.5, + "valid_targets_min": 912 + }, + { + "epoch": 4.851157662624035, + "grad_norm": 0.5478924490850784, + "learning_rate": 1.0429939009431229e-05, + "loss": 0.1076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04804785177111626, + "step": 8800, + "valid_targets_mean": 2147.6, + "valid_targets_min": 1035 + }, + { + "epoch": 4.853914002205071, + "grad_norm": 0.6044070631700452, + "learning_rate": 1.0405809203804713e-05, + "loss": 0.104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08263874053955078, + "step": 8805, + "valid_targets_mean": 1750.5, + "valid_targets_min": 926 + }, + { + "epoch": 4.856670341786108, + "grad_norm": 0.4533948791390458, + "learning_rate": 1.0381697524411817e-05, + "loss": 0.0794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032628294080495834, + "step": 8810, + "valid_targets_mean": 2832.5, + "valid_targets_min": 703 + }, + { + "epoch": 4.859426681367144, + "grad_norm": 0.35903866006272533, + "learning_rate": 1.0357604016806562e-05, + "loss": 0.0845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03680954873561859, + "step": 8815, + "valid_targets_mean": 3462.8, + "valid_targets_min": 2841 + }, + { + "epoch": 4.862183020948181, + "grad_norm": 0.3744313062115718, + "learning_rate": 1.0333528726508628e-05, + "loss": 0.0837, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030485838651657104, + "step": 8820, + "valid_targets_mean": 2462.2, + "valid_targets_min": 745 + }, + { + "epoch": 4.864939360529217, + "grad_norm": 0.36689752156988087, + "learning_rate": 1.0309471699003288e-05, + "loss": 0.0726, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037025950849056244, + "step": 8825, + "valid_targets_mean": 4582.6, + "valid_targets_min": 1048 + }, + { + "epoch": 4.867695700110254, + "grad_norm": 0.3299457537340771, + "learning_rate": 1.0285432979741295e-05, + "loss": 0.0669, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0355803407728672, + "step": 8830, + "valid_targets_mean": 3956.1, + "valid_targets_min": 2357 + }, + { + "epoch": 4.87045203969129, + "grad_norm": 0.32639559106560745, + "learning_rate": 1.0261412614138839e-05, + "loss": 0.0786, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033042944967746735, + "step": 8835, + "valid_targets_mean": 3944.2, + "valid_targets_min": 1850 + }, + { + "epoch": 4.873208379272326, + "grad_norm": 0.3849346957518695, + "learning_rate": 1.0237410647577415e-05, + "loss": 0.0891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03437165170907974, + "step": 8840, + "valid_targets_mean": 3158.4, + "valid_targets_min": 889 + }, + { + "epoch": 4.875964718853362, + "grad_norm": 0.4483665090781295, + "learning_rate": 1.0213427125403756e-05, + "loss": 0.0868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04097796976566315, + "step": 8845, + "valid_targets_mean": 2407.2, + "valid_targets_min": 749 + }, + { + "epoch": 4.878721058434399, + "grad_norm": 0.29953890171807135, + "learning_rate": 1.0189462092929756e-05, + "loss": 0.0878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029940592125058174, + "step": 8850, + "valid_targets_mean": 4191.4, + "valid_targets_min": 3925 + }, + { + "epoch": 4.881477398015435, + "grad_norm": 0.43759458205001994, + "learning_rate": 1.016551559543237e-05, + "loss": 0.0713, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04331284388899803, + "step": 8855, + "valid_targets_mean": 3605.1, + "valid_targets_min": 1567 + }, + { + "epoch": 4.884233737596472, + "grad_norm": 0.46413962259492175, + "learning_rate": 1.0141587678153542e-05, + "loss": 0.0844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03590533137321472, + "step": 8860, + "valid_targets_mean": 2819.0, + "valid_targets_min": 902 + }, + { + "epoch": 4.886990077177508, + "grad_norm": 0.3367619018705965, + "learning_rate": 1.0117678386300107e-05, + "loss": 0.0739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028451543301343918, + "step": 8865, + "valid_targets_mean": 3424.2, + "valid_targets_min": 974 + }, + { + "epoch": 4.889746416758545, + "grad_norm": 0.5988172253518782, + "learning_rate": 1.0093787765043715e-05, + "loss": 0.0811, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045291393995285034, + "step": 8870, + "valid_targets_mean": 1211.8, + "valid_targets_min": 569 + }, + { + "epoch": 4.892502756339581, + "grad_norm": 0.518239458832006, + "learning_rate": 1.0069915859520726e-05, + "loss": 0.0771, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05555427074432373, + "step": 8875, + "valid_targets_mean": 2852.6, + "valid_targets_min": 1015 + }, + { + "epoch": 4.8952590959206175, + "grad_norm": 0.511863076320767, + "learning_rate": 1.0046062714832176e-05, + "loss": 0.0784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048268940299749374, + "step": 8880, + "valid_targets_mean": 1880.9, + "valid_targets_min": 677 + }, + { + "epoch": 4.898015435501653, + "grad_norm": 0.3964482620661039, + "learning_rate": 1.0022228376043626e-05, + "loss": 0.0929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03942461684346199, + "step": 8885, + "valid_targets_mean": 2526.8, + "valid_targets_min": 1010 + }, + { + "epoch": 4.90077177508269, + "grad_norm": 0.5956960131426524, + "learning_rate": 9.998412888185117e-06, + "loss": 0.0782, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04596507549285889, + "step": 8890, + "valid_targets_mean": 1515.2, + "valid_targets_min": 700 + }, + { + "epoch": 4.903528114663727, + "grad_norm": 0.3830741429673215, + "learning_rate": 9.974616296251067e-06, + "loss": 0.0901, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028072655200958252, + "step": 8895, + "valid_targets_mean": 2491.1, + "valid_targets_min": 619 + }, + { + "epoch": 4.906284454244763, + "grad_norm": 0.7559060270645087, + "learning_rate": 9.950838645200209e-06, + "loss": 0.1003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03900078684091568, + "step": 8900, + "valid_targets_mean": 3964.8, + "valid_targets_min": 3561 + }, + { + "epoch": 4.909040793825799, + "grad_norm": 0.3937004324036139, + "learning_rate": 9.927079979955479e-06, + "loss": 0.0812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033228084444999695, + "step": 8905, + "valid_targets_mean": 2793.2, + "valid_targets_min": 809 + }, + { + "epoch": 4.911797133406836, + "grad_norm": 0.43359557688875355, + "learning_rate": 9.903340345403947e-06, + "loss": 0.0701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044144004583358765, + "step": 8910, + "valid_targets_mean": 3057.9, + "valid_targets_min": 868 + }, + { + "epoch": 4.914553472987873, + "grad_norm": 0.48682997227787683, + "learning_rate": 9.87961978639673e-06, + "loss": 0.0848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07439523190259933, + "step": 8915, + "valid_targets_mean": 2886.0, + "valid_targets_min": 1315 + }, + { + "epoch": 4.9173098125689085, + "grad_norm": 0.3981160319449798, + "learning_rate": 9.855918347748905e-06, + "loss": 0.0937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03205881267786026, + "step": 8920, + "valid_targets_mean": 3051.6, + "valid_targets_min": 1103 + }, + { + "epoch": 4.9200661521499445, + "grad_norm": 0.383656794338573, + "learning_rate": 9.832236074239415e-06, + "loss": 0.0618, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029673775658011436, + "step": 8925, + "valid_targets_mean": 2638.6, + "valid_targets_min": 680 + }, + { + "epoch": 4.922822491730981, + "grad_norm": 0.435961288061073, + "learning_rate": 9.80857301061102e-06, + "loss": 0.0833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03674326092004776, + "step": 8930, + "valid_targets_mean": 2283.4, + "valid_targets_min": 667 + }, + { + "epoch": 4.925578831312018, + "grad_norm": 0.3736864129529646, + "learning_rate": 9.784929201570166e-06, + "loss": 0.0737, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03592894226312637, + "step": 8935, + "valid_targets_mean": 4612.6, + "valid_targets_min": 3668 + }, + { + "epoch": 4.928335170893054, + "grad_norm": 0.3633197384997168, + "learning_rate": 9.761304691786921e-06, + "loss": 0.0645, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03248334303498268, + "step": 8940, + "valid_targets_mean": 3685.1, + "valid_targets_min": 1236 + }, + { + "epoch": 4.93109151047409, + "grad_norm": 0.3651484279250123, + "learning_rate": 9.737699525894902e-06, + "loss": 0.0867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03726832568645477, + "step": 8945, + "valid_targets_mean": 4063.4, + "valid_targets_min": 3222 + }, + { + "epoch": 4.933847850055127, + "grad_norm": 0.48881730120618344, + "learning_rate": 9.714113748491167e-06, + "loss": 0.0727, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0436459556221962, + "step": 8950, + "valid_targets_mean": 1614.5, + "valid_targets_min": 676 + }, + { + "epoch": 4.936604189636164, + "grad_norm": 0.4287484486281858, + "learning_rate": 9.69054740413616e-06, + "loss": 0.101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03877885267138481, + "step": 8955, + "valid_targets_mean": 2698.2, + "valid_targets_min": 561 + }, + { + "epoch": 4.9393605292172, + "grad_norm": 0.26495741012757223, + "learning_rate": 9.667000537353593e-06, + "loss": 0.0877, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.01759764924645424, + "step": 8960, + "valid_targets_mean": 3602.9, + "valid_targets_min": 924 + }, + { + "epoch": 4.9421168687982355, + "grad_norm": 0.5583185829176509, + "learning_rate": 9.643473192630391e-06, + "loss": 0.0929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08261610567569733, + "step": 8965, + "valid_targets_mean": 3460.0, + "valid_targets_min": 3039 + }, + { + "epoch": 4.944873208379272, + "grad_norm": 0.39952921191984997, + "learning_rate": 9.61996541441659e-06, + "loss": 0.0736, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03683662787079811, + "step": 8970, + "valid_targets_mean": 3788.9, + "valid_targets_min": 2613 + }, + { + "epoch": 4.947629547960309, + "grad_norm": 0.6765951084941558, + "learning_rate": 9.596477247125253e-06, + "loss": 0.0818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03982534632086754, + "step": 8975, + "valid_targets_mean": 865.4, + "valid_targets_min": 690 + }, + { + "epoch": 4.950385887541345, + "grad_norm": 0.40601846932198804, + "learning_rate": 9.57300873513242e-06, + "loss": 0.0752, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03990977257490158, + "step": 8980, + "valid_targets_mean": 3928.9, + "valid_targets_min": 3584 + }, + { + "epoch": 4.953142227122381, + "grad_norm": 0.85010395625867, + "learning_rate": 9.549559922776968e-06, + "loss": 0.1011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05125194042921066, + "step": 8985, + "valid_targets_mean": 1228.5, + "valid_targets_min": 540 + }, + { + "epoch": 4.955898566703418, + "grad_norm": 0.5914378785650174, + "learning_rate": 9.52613085436056e-06, + "loss": 0.0985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0539291575551033, + "step": 8990, + "valid_targets_mean": 2018.4, + "valid_targets_min": 500 + }, + { + "epoch": 4.958654906284455, + "grad_norm": 0.4862879599122032, + "learning_rate": 9.502721574147567e-06, + "loss": 0.0808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05114535614848137, + "step": 8995, + "valid_targets_mean": 2198.2, + "valid_targets_min": 622 + }, + { + "epoch": 4.961411245865491, + "grad_norm": 0.44291136278395676, + "learning_rate": 9.479332126364966e-06, + "loss": 0.0784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03350115567445755, + "step": 9000, + "valid_targets_mean": 3576.6, + "valid_targets_min": 2812 + }, + { + "epoch": 4.964167585446527, + "grad_norm": 0.427755903986499, + "learning_rate": 9.455962555202267e-06, + "loss": 0.0777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047360315918922424, + "step": 9005, + "valid_targets_mean": 3544.9, + "valid_targets_min": 2977 + }, + { + "epoch": 4.966923925027563, + "grad_norm": 0.38664270725392136, + "learning_rate": 9.432612904811429e-06, + "loss": 0.0708, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03388415277004242, + "step": 9010, + "valid_targets_mean": 2623.5, + "valid_targets_min": 852 + }, + { + "epoch": 4.9696802646086, + "grad_norm": 0.9203994634513053, + "learning_rate": 9.409283219306773e-06, + "loss": 0.0781, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04919441044330597, + "step": 9015, + "valid_targets_mean": 830.8, + "valid_targets_min": 603 + }, + { + "epoch": 4.972436604189636, + "grad_norm": 0.4026831137391381, + "learning_rate": 9.385973542764895e-06, + "loss": 0.0802, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029653996229171753, + "step": 9020, + "valid_targets_mean": 2720.6, + "valid_targets_min": 747 + }, + { + "epoch": 4.975192943770672, + "grad_norm": 0.5894575425664907, + "learning_rate": 9.362683919224606e-06, + "loss": 0.0725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04799339920282364, + "step": 9025, + "valid_targets_mean": 2823.8, + "valid_targets_min": 850 + }, + { + "epoch": 4.977949283351709, + "grad_norm": 0.458195658023055, + "learning_rate": 9.339414392686816e-06, + "loss": 0.072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04554351791739464, + "step": 9030, + "valid_targets_mean": 3792.5, + "valid_targets_min": 2890 + }, + { + "epoch": 4.980705622932746, + "grad_norm": 0.3452668859071227, + "learning_rate": 9.316165007114469e-06, + "loss": 0.0816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029914192855358124, + "step": 9035, + "valid_targets_mean": 3551.2, + "valid_targets_min": 482 + }, + { + "epoch": 4.983461962513782, + "grad_norm": 0.2897512792349616, + "learning_rate": 9.29293580643245e-06, + "loss": 0.068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03305509686470032, + "step": 9040, + "valid_targets_mean": 4891.9, + "valid_targets_min": 2535 + }, + { + "epoch": 4.986218302094818, + "grad_norm": 0.3466938108654639, + "learning_rate": 9.269726834527524e-06, + "loss": 0.0738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038864150643348694, + "step": 9045, + "valid_targets_mean": 4404.5, + "valid_targets_min": 3353 + }, + { + "epoch": 4.9889746416758545, + "grad_norm": 0.43005406490105436, + "learning_rate": 9.246538135248227e-06, + "loss": 0.0748, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04833151027560234, + "step": 9050, + "valid_targets_mean": 2697.8, + "valid_targets_min": 867 + }, + { + "epoch": 4.991730981256891, + "grad_norm": 0.36697707037513166, + "learning_rate": 9.223369752404794e-06, + "loss": 0.159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033151399344205856, + "step": 9055, + "valid_targets_mean": 2442.9, + "valid_targets_min": 1009 + }, + { + "epoch": 4.994487320837927, + "grad_norm": 0.3718043826415328, + "learning_rate": 9.200221729769084e-06, + "loss": 0.0737, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05005926638841629, + "step": 9060, + "valid_targets_mean": 2321.9, + "valid_targets_min": 1144 + }, + { + "epoch": 4.997243660418963, + "grad_norm": 0.3485090002597888, + "learning_rate": 9.17709411107448e-06, + "loss": 0.0756, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02902999147772789, + "step": 9065, + "valid_targets_mean": 3562.4, + "valid_targets_min": 2942 + }, + { + "epoch": 5.0, + "grad_norm": 0.40727703821779115, + "learning_rate": 9.153986940015817e-06, + "loss": 0.1049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045864395797252655, + "step": 9070, + "valid_targets_mean": 4458.2, + "valid_targets_min": 1268 + }, + { + "epoch": 5.002756339581037, + "grad_norm": 0.6667589625692739, + "learning_rate": 9.130900260249315e-06, + "loss": 0.1893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08985529839992523, + "step": 9075, + "valid_targets_mean": 7359.2, + "valid_targets_min": 5595 + }, + { + "epoch": 5.005512679162073, + "grad_norm": 0.6476852491051197, + "learning_rate": 9.10783411539246e-06, + "loss": 0.1812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04788250848650932, + "step": 9080, + "valid_targets_mean": 1287.1, + "valid_targets_min": 420 + }, + { + "epoch": 5.00826901874311, + "grad_norm": 0.41384496657809094, + "learning_rate": 9.084788549023951e-06, + "loss": 0.173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09059896320104599, + "step": 9085, + "valid_targets_mean": 8512.8, + "valid_targets_min": 6729 + }, + { + "epoch": 5.0110253583241455, + "grad_norm": 0.3771114645810749, + "learning_rate": 9.061763604683603e-06, + "loss": 0.1688, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07927650213241577, + "step": 9090, + "valid_targets_mean": 7552.0, + "valid_targets_min": 6145 + }, + { + "epoch": 5.013781697905182, + "grad_norm": 0.3680994752271587, + "learning_rate": 9.03875932587227e-06, + "loss": 0.1702, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07695163786411285, + "step": 9095, + "valid_targets_mean": 7301.4, + "valid_targets_min": 4807 + }, + { + "epoch": 5.016538037486218, + "grad_norm": 0.4013679164091624, + "learning_rate": 9.015775756051774e-06, + "loss": 0.1787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09161560237407684, + "step": 9100, + "valid_targets_mean": 6809.0, + "valid_targets_min": 5015 + }, + { + "epoch": 5.019294377067255, + "grad_norm": 0.39751249698106644, + "learning_rate": 8.992812938644798e-06, + "loss": 0.1825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09317012131214142, + "step": 9105, + "valid_targets_mean": 7217.0, + "valid_targets_min": 5541 + }, + { + "epoch": 5.022050716648291, + "grad_norm": 0.3799380663970374, + "learning_rate": 8.969870917034821e-06, + "loss": 0.1688, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07660777866840363, + "step": 9110, + "valid_targets_mean": 6205.1, + "valid_targets_min": 5580 + }, + { + "epoch": 5.024807056229328, + "grad_norm": 0.846149504151175, + "learning_rate": 8.946949734566037e-06, + "loss": 0.1635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.024922046810388565, + "step": 9115, + "valid_targets_mean": 598.0, + "valid_targets_min": 151 + }, + { + "epoch": 5.027563395810364, + "grad_norm": 0.39144709477826745, + "learning_rate": 8.92404943454326e-06, + "loss": 0.1562, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07953938841819763, + "step": 9120, + "valid_targets_mean": 7270.8, + "valid_targets_min": 6125 + }, + { + "epoch": 5.030319735391401, + "grad_norm": 0.38246013982953125, + "learning_rate": 8.901170060231873e-06, + "loss": 0.1721, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0798879936337471, + "step": 9125, + "valid_targets_mean": 6108.9, + "valid_targets_min": 4623 + }, + { + "epoch": 5.033076074972437, + "grad_norm": 0.4134004420138146, + "learning_rate": 8.878311654857697e-06, + "loss": 0.1638, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08560197055339813, + "step": 9130, + "valid_targets_mean": 6571.6, + "valid_targets_min": 4850 + }, + { + "epoch": 5.035832414553473, + "grad_norm": 0.40587067658691905, + "learning_rate": 8.855474261606948e-06, + "loss": 0.1661, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08410146832466125, + "step": 9135, + "valid_targets_mean": 7984.4, + "valid_targets_min": 6312 + }, + { + "epoch": 5.038588754134509, + "grad_norm": 0.3627107006456358, + "learning_rate": 8.832657923626149e-06, + "loss": 0.1579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07392202317714691, + "step": 9140, + "valid_targets_mean": 6273.1, + "valid_targets_min": 4590 + }, + { + "epoch": 5.041345093715546, + "grad_norm": 0.37840799388169666, + "learning_rate": 8.809862684022032e-06, + "loss": 0.1668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08733304589986801, + "step": 9145, + "valid_targets_mean": 7860.2, + "valid_targets_min": 5453 + }, + { + "epoch": 5.044101433296582, + "grad_norm": 0.5351628718793552, + "learning_rate": 8.787088585861481e-06, + "loss": 0.1659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07586371898651123, + "step": 9150, + "valid_targets_mean": 6988.6, + "valid_targets_min": 4928 + }, + { + "epoch": 5.046857772877619, + "grad_norm": 0.4084335160249153, + "learning_rate": 8.764335672171428e-06, + "loss": 0.1626, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07932071387767792, + "step": 9155, + "valid_targets_mean": 6495.6, + "valid_targets_min": 5367 + }, + { + "epoch": 5.049614112458655, + "grad_norm": 0.4446692085663556, + "learning_rate": 8.74160398593878e-06, + "loss": 0.1686, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08171440660953522, + "step": 9160, + "valid_targets_mean": 4968.1, + "valid_targets_min": 615 + }, + { + "epoch": 5.052370452039692, + "grad_norm": 0.4104255975766886, + "learning_rate": 8.718893570110343e-06, + "loss": 0.183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09284122288227081, + "step": 9165, + "valid_targets_mean": 7189.9, + "valid_targets_min": 5732 + }, + { + "epoch": 5.055126791620728, + "grad_norm": 0.378604123916646, + "learning_rate": 8.696204467592749e-06, + "loss": 0.1485, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0832129493355751, + "step": 9170, + "valid_targets_mean": 7969.4, + "valid_targets_min": 5870 + }, + { + "epoch": 5.0578831312017645, + "grad_norm": 0.36093284901823397, + "learning_rate": 8.673536721252342e-06, + "loss": 0.1538, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0657939612865448, + "step": 9175, + "valid_targets_mean": 7348.0, + "valid_targets_min": 4983 + }, + { + "epoch": 5.0606394707828, + "grad_norm": 0.396455454766918, + "learning_rate": 8.65089037391513e-06, + "loss": 0.1633, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06843159347772598, + "step": 9180, + "valid_targets_mean": 6763.5, + "valid_targets_min": 5251 + }, + { + "epoch": 5.063395810363837, + "grad_norm": 0.4064164162989706, + "learning_rate": 8.62826546836669e-06, + "loss": 0.1673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09268009662628174, + "step": 9185, + "valid_targets_mean": 6626.4, + "valid_targets_min": 4332 + }, + { + "epoch": 5.066152149944873, + "grad_norm": 0.37686383962708625, + "learning_rate": 8.605662047352088e-06, + "loss": 0.1667, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07682083547115326, + "step": 9190, + "valid_targets_mean": 7151.6, + "valid_targets_min": 4434 + }, + { + "epoch": 5.06890848952591, + "grad_norm": 0.4257154086128477, + "learning_rate": 8.583080153575797e-06, + "loss": 0.1654, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09809395670890808, + "step": 9195, + "valid_targets_mean": 7592.0, + "valid_targets_min": 5705 + }, + { + "epoch": 5.071664829106946, + "grad_norm": 0.36547905684742354, + "learning_rate": 8.560519829701627e-06, + "loss": 0.1461, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06883938610553741, + "step": 9200, + "valid_targets_mean": 8727.9, + "valid_targets_min": 5721 + }, + { + "epoch": 5.074421168687983, + "grad_norm": 0.43165897560806804, + "learning_rate": 8.537981118352627e-06, + "loss": 0.1465, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07676833868026733, + "step": 9205, + "valid_targets_mean": 6314.9, + "valid_targets_min": 4517 + }, + { + "epoch": 5.077177508269019, + "grad_norm": 0.39943635839930663, + "learning_rate": 8.515464062111017e-06, + "loss": 0.1433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08061368018388748, + "step": 9210, + "valid_targets_mean": 7161.6, + "valid_targets_min": 4531 + }, + { + "epoch": 5.0799338478500555, + "grad_norm": 0.7088118380312447, + "learning_rate": 8.492968703518099e-06, + "loss": 0.1678, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05163313075900078, + "step": 9215, + "valid_targets_mean": 1098.8, + "valid_targets_min": 160 + }, + { + "epoch": 5.0826901874310915, + "grad_norm": 0.39892240358360254, + "learning_rate": 8.470495085074201e-06, + "loss": 0.1514, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07522940635681152, + "step": 9220, + "valid_targets_mean": 6988.4, + "valid_targets_min": 5392 + }, + { + "epoch": 5.085446527012128, + "grad_norm": 0.41525583064104915, + "learning_rate": 8.448043249238557e-06, + "loss": 0.1717, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08822352439165115, + "step": 9225, + "valid_targets_mean": 7712.9, + "valid_targets_min": 5297 + }, + { + "epoch": 5.088202866593164, + "grad_norm": 0.41659201195117157, + "learning_rate": 8.425613238429256e-06, + "loss": 0.1746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08296297490596771, + "step": 9230, + "valid_targets_mean": 6396.9, + "valid_targets_min": 5536 + }, + { + "epoch": 5.090959206174201, + "grad_norm": 0.42973577089470666, + "learning_rate": 8.403205095023148e-06, + "loss": 0.1707, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08451573550701141, + "step": 9235, + "valid_targets_mean": 6143.8, + "valid_targets_min": 3434 + }, + { + "epoch": 5.093715545755237, + "grad_norm": 0.3832206717555118, + "learning_rate": 8.380818861355775e-06, + "loss": 0.1651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0859198123216629, + "step": 9240, + "valid_targets_mean": 7182.1, + "valid_targets_min": 5349 + }, + { + "epoch": 5.096471885336274, + "grad_norm": 0.3837290554466305, + "learning_rate": 8.358454579721283e-06, + "loss": 0.1705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0866212248802185, + "step": 9245, + "valid_targets_mean": 7570.5, + "valid_targets_min": 5209 + }, + { + "epoch": 5.09922822491731, + "grad_norm": 0.398459035024381, + "learning_rate": 8.33611229237234e-06, + "loss": 0.1652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09082755446434021, + "step": 9250, + "valid_targets_mean": 7069.0, + "valid_targets_min": 5581 + }, + { + "epoch": 5.101984564498347, + "grad_norm": 0.4168849834378292, + "learning_rate": 8.313792041520066e-06, + "loss": 0.1687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08924612402915955, + "step": 9255, + "valid_targets_mean": 6925.6, + "valid_targets_min": 5504 + }, + { + "epoch": 5.1047409040793825, + "grad_norm": 0.4027660633492839, + "learning_rate": 8.29149386933394e-06, + "loss": 0.1697, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08666248619556427, + "step": 9260, + "valid_targets_mean": 6716.9, + "valid_targets_min": 5332 + }, + { + "epoch": 5.107497243660419, + "grad_norm": 0.44707305024740124, + "learning_rate": 8.269217817941747e-06, + "loss": 0.1702, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08148002624511719, + "step": 9265, + "valid_targets_mean": 6241.8, + "valid_targets_min": 5406 + }, + { + "epoch": 5.110253583241455, + "grad_norm": 0.38596158435326855, + "learning_rate": 8.246963929429454e-06, + "loss": 0.1658, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07789814472198486, + "step": 9270, + "valid_targets_mean": 6387.8, + "valid_targets_min": 4847 + }, + { + "epoch": 5.113009922822492, + "grad_norm": 0.4119558004059517, + "learning_rate": 8.224732245841176e-06, + "loss": 0.1648, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07784758508205414, + "step": 9275, + "valid_targets_mean": 6457.6, + "valid_targets_min": 4288 + }, + { + "epoch": 5.115766262403528, + "grad_norm": 0.4412084857513274, + "learning_rate": 8.202522809179065e-06, + "loss": 0.1355, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09699342399835587, + "step": 9280, + "valid_targets_mean": 6793.1, + "valid_targets_min": 5173 + }, + { + "epoch": 5.118522601984565, + "grad_norm": 0.4350142664882942, + "learning_rate": 8.180335661403242e-06, + "loss": 0.1634, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0765116959810257, + "step": 9285, + "valid_targets_mean": 6697.4, + "valid_targets_min": 5168 + }, + { + "epoch": 5.121278941565601, + "grad_norm": 0.4129119935320094, + "learning_rate": 8.158170844431725e-06, + "loss": 0.1626, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07860542088747025, + "step": 9290, + "valid_targets_mean": 6062.0, + "valid_targets_min": 5270 + }, + { + "epoch": 5.124035281146638, + "grad_norm": 0.41232589500808214, + "learning_rate": 8.136028400140337e-06, + "loss": 0.1681, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08677247166633606, + "step": 9295, + "valid_targets_mean": 7110.1, + "valid_targets_min": 5253 + }, + { + "epoch": 5.126791620727674, + "grad_norm": 0.3637833051325588, + "learning_rate": 8.113908370362633e-06, + "loss": 0.1596, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07655549794435501, + "step": 9300, + "valid_targets_mean": 7395.9, + "valid_targets_min": 5248 + }, + { + "epoch": 5.12954796030871, + "grad_norm": 0.4685590906750839, + "learning_rate": 8.091810796889825e-06, + "loss": 0.1671, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10588568449020386, + "step": 9305, + "valid_targets_mean": 7033.5, + "valid_targets_min": 5354 + }, + { + "epoch": 5.132304299889746, + "grad_norm": 0.39955819891517624, + "learning_rate": 8.069735721470685e-06, + "loss": 0.1665, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08600588142871857, + "step": 9310, + "valid_targets_mean": 6798.8, + "valid_targets_min": 4532 + }, + { + "epoch": 5.135060639470783, + "grad_norm": 0.5183579072686708, + "learning_rate": 8.047683185811501e-06, + "loss": 0.1438, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037983208894729614, + "step": 9315, + "valid_targets_mean": 2045.0, + "valid_targets_min": 164 + }, + { + "epoch": 5.137816979051819, + "grad_norm": 0.4127782006782886, + "learning_rate": 8.025653231575963e-06, + "loss": 0.1483, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07880507409572601, + "step": 9320, + "valid_targets_mean": 6723.1, + "valid_targets_min": 5164 + }, + { + "epoch": 5.140573318632856, + "grad_norm": 0.4498385834903266, + "learning_rate": 8.003645900385097e-06, + "loss": 0.1624, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08329832553863525, + "step": 9325, + "valid_targets_mean": 5954.1, + "valid_targets_min": 4996 + }, + { + "epoch": 5.143329658213892, + "grad_norm": 0.4297777136487167, + "learning_rate": 7.981661233817196e-06, + "loss": 0.1666, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07759891450405121, + "step": 9330, + "valid_targets_mean": 5932.0, + "valid_targets_min": 5205 + }, + { + "epoch": 5.146085997794929, + "grad_norm": 0.420600401912816, + "learning_rate": 7.959699273407722e-06, + "loss": 0.1619, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0843975767493248, + "step": 9335, + "valid_targets_mean": 6288.4, + "valid_targets_min": 5380 + }, + { + "epoch": 5.148842337375965, + "grad_norm": 0.38310068851442286, + "learning_rate": 7.937760060649247e-06, + "loss": 0.1534, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06666086614131927, + "step": 9340, + "valid_targets_mean": 7095.9, + "valid_targets_min": 5048 + }, + { + "epoch": 5.1515986769570015, + "grad_norm": 0.41015279751247696, + "learning_rate": 7.915843636991362e-06, + "loss": 0.1644, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0799093246459961, + "step": 9345, + "valid_targets_mean": 6093.5, + "valid_targets_min": 4489 + }, + { + "epoch": 5.154355016538037, + "grad_norm": 0.8374200538355047, + "learning_rate": 7.893950043840605e-06, + "loss": 0.1594, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09426723420619965, + "step": 9350, + "valid_targets_mean": 1659.9, + "valid_targets_min": 857 + }, + { + "epoch": 5.157111356119074, + "grad_norm": 0.7336875486569768, + "learning_rate": 7.87207932256038e-06, + "loss": 0.1429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07051660865545273, + "step": 9355, + "valid_targets_mean": 1260.4, + "valid_targets_min": 686 + }, + { + "epoch": 5.15986769570011, + "grad_norm": 0.7253505451896352, + "learning_rate": 7.85023151447087e-06, + "loss": 0.1363, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05211283266544342, + "step": 9360, + "valid_targets_mean": 1226.2, + "valid_targets_min": 676 + }, + { + "epoch": 5.162624035281147, + "grad_norm": 0.7693140435652747, + "learning_rate": 7.82840666084899e-06, + "loss": 0.1285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05437218397855759, + "step": 9365, + "valid_targets_mean": 1129.9, + "valid_targets_min": 661 + }, + { + "epoch": 5.165380374862183, + "grad_norm": 0.7522893755108245, + "learning_rate": 7.806604802928272e-06, + "loss": 0.1288, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0668160617351532, + "step": 9370, + "valid_targets_mean": 1631.2, + "valid_targets_min": 520 + }, + { + "epoch": 5.16813671444322, + "grad_norm": 0.8394250766879807, + "learning_rate": 7.7848259818988e-06, + "loss": 0.1237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06145526468753815, + "step": 9375, + "valid_targets_mean": 1091.9, + "valid_targets_min": 472 + }, + { + "epoch": 5.170893054024256, + "grad_norm": 0.8078850687211672, + "learning_rate": 7.76307023890714e-06, + "loss": 0.1339, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05336114391684532, + "step": 9380, + "valid_targets_mean": 1244.9, + "valid_targets_min": 497 + }, + { + "epoch": 5.1736493936052925, + "grad_norm": 0.8326080724402387, + "learning_rate": 7.741337615056259e-06, + "loss": 0.1164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06387875974178314, + "step": 9385, + "valid_targets_mean": 1297.4, + "valid_targets_min": 453 + }, + { + "epoch": 5.1764057331863285, + "grad_norm": 0.7883513353638261, + "learning_rate": 7.71962815140544e-06, + "loss": 0.1415, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05788443237543106, + "step": 9390, + "valid_targets_mean": 1147.8, + "valid_targets_min": 501 + }, + { + "epoch": 5.179162072767365, + "grad_norm": 0.8204896145253705, + "learning_rate": 7.697941888970213e-06, + "loss": 0.1239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043426305055618286, + "step": 9395, + "valid_targets_mean": 1010.0, + "valid_targets_min": 580 + }, + { + "epoch": 5.181918412348401, + "grad_norm": 0.8344584151387274, + "learning_rate": 7.67627886872227e-06, + "loss": 0.1261, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07460416853427887, + "step": 9400, + "valid_targets_mean": 1627.9, + "valid_targets_min": 698 + }, + { + "epoch": 5.184674751929438, + "grad_norm": 0.8228048413313692, + "learning_rate": 7.654639131589394e-06, + "loss": 0.1236, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05188912898302078, + "step": 9405, + "valid_targets_mean": 1142.4, + "valid_targets_min": 636 + }, + { + "epoch": 5.187431091510474, + "grad_norm": 0.8997288091061901, + "learning_rate": 7.633022718455387e-06, + "loss": 0.1265, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06674276292324066, + "step": 9410, + "valid_targets_mean": 1073.8, + "valid_targets_min": 568 + }, + { + "epoch": 5.190187431091511, + "grad_norm": 0.7700264238928992, + "learning_rate": 7.611429670159976e-06, + "loss": 0.1114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06060558930039406, + "step": 9415, + "valid_targets_mean": 1611.5, + "valid_targets_min": 982 + }, + { + "epoch": 5.192943770672547, + "grad_norm": 0.7908337424791627, + "learning_rate": 7.589860027498746e-06, + "loss": 0.1109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05109712481498718, + "step": 9420, + "valid_targets_mean": 1190.0, + "valid_targets_min": 921 + }, + { + "epoch": 5.195700110253584, + "grad_norm": 0.9416398686947677, + "learning_rate": 7.568313831223062e-06, + "loss": 0.1167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05768633633852005, + "step": 9425, + "valid_targets_mean": 1367.0, + "valid_targets_min": 559 + }, + { + "epoch": 5.1984564498346195, + "grad_norm": 0.8524140054315525, + "learning_rate": 7.546791122039994e-06, + "loss": 0.118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07685214281082153, + "step": 9430, + "valid_targets_mean": 1773.5, + "valid_targets_min": 609 + }, + { + "epoch": 5.201212789415656, + "grad_norm": 0.9143828376024539, + "learning_rate": 7.525291940612236e-06, + "loss": 0.122, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07212072610855103, + "step": 9435, + "valid_targets_mean": 1784.5, + "valid_targets_min": 865 + }, + { + "epoch": 5.203969128996692, + "grad_norm": 0.8035749882664905, + "learning_rate": 7.503816327558036e-06, + "loss": 0.1182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05386310815811157, + "step": 9440, + "valid_targets_mean": 1142.9, + "valid_targets_min": 858 + }, + { + "epoch": 5.206725468577729, + "grad_norm": 0.9236017942462253, + "learning_rate": 7.482364323451106e-06, + "loss": 0.1231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06482227891683578, + "step": 9445, + "valid_targets_mean": 1556.0, + "valid_targets_min": 461 + }, + { + "epoch": 5.209481808158765, + "grad_norm": 0.8898360630929804, + "learning_rate": 7.46093596882056e-06, + "loss": 0.1232, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04969433695077896, + "step": 9450, + "valid_targets_mean": 882.1, + "valid_targets_min": 476 + }, + { + "epoch": 5.212238147739802, + "grad_norm": 0.8416895647798167, + "learning_rate": 7.439531304150824e-06, + "loss": 0.1258, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06067527085542679, + "step": 9455, + "valid_targets_mean": 1290.1, + "valid_targets_min": 469 + }, + { + "epoch": 5.214994487320838, + "grad_norm": 0.8129194366018907, + "learning_rate": 7.4181503698815845e-06, + "loss": 0.1197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04191707819700241, + "step": 9460, + "valid_targets_mean": 1085.0, + "valid_targets_min": 644 + }, + { + "epoch": 5.217750826901875, + "grad_norm": 0.8054745003162227, + "learning_rate": 7.396793206407677e-06, + "loss": 0.1098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06924641132354736, + "step": 9465, + "valid_targets_mean": 1322.4, + "valid_targets_min": 673 + }, + { + "epoch": 5.220507166482911, + "grad_norm": 0.9264657024201123, + "learning_rate": 7.375459854079034e-06, + "loss": 0.1179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05807020515203476, + "step": 9470, + "valid_targets_mean": 1386.5, + "valid_targets_min": 574 + }, + { + "epoch": 5.223263506063947, + "grad_norm": 0.9350118345080792, + "learning_rate": 7.354150353200596e-06, + "loss": 0.1241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06641645729541779, + "step": 9475, + "valid_targets_mean": 1325.2, + "valid_targets_min": 513 + }, + { + "epoch": 5.226019845644983, + "grad_norm": 0.8320937104135182, + "learning_rate": 7.332864744032246e-06, + "loss": 0.1194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04302702471613884, + "step": 9480, + "valid_targets_mean": 1024.5, + "valid_targets_min": 670 + }, + { + "epoch": 5.22877618522602, + "grad_norm": 0.8286187672576933, + "learning_rate": 7.31160306678873e-06, + "loss": 0.1238, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060765817761421204, + "step": 9485, + "valid_targets_mean": 1211.0, + "valid_targets_min": 583 + }, + { + "epoch": 5.231532524807056, + "grad_norm": 0.7930339858162399, + "learning_rate": 7.290365361639577e-06, + "loss": 0.1174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05159320682287216, + "step": 9490, + "valid_targets_mean": 1101.4, + "valid_targets_min": 503 + }, + { + "epoch": 5.234288864388093, + "grad_norm": 0.8271924101506325, + "learning_rate": 7.269151668709024e-06, + "loss": 0.1181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0436023473739624, + "step": 9495, + "valid_targets_mean": 995.1, + "valid_targets_min": 642 + }, + { + "epoch": 5.237045203969129, + "grad_norm": 0.8212895414076674, + "learning_rate": 7.247962028075946e-06, + "loss": 0.1156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05201122164726257, + "step": 9500, + "valid_targets_mean": 1149.6, + "valid_targets_min": 530 + }, + { + "epoch": 5.239801543550166, + "grad_norm": 0.7643368234201882, + "learning_rate": 7.226796479773763e-06, + "loss": 0.1166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04685389995574951, + "step": 9505, + "valid_targets_mean": 1142.5, + "valid_targets_min": 805 + }, + { + "epoch": 5.242557883131202, + "grad_norm": 0.8745132982495796, + "learning_rate": 7.205655063790408e-06, + "loss": 0.1185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052907612174749374, + "step": 9510, + "valid_targets_mean": 1230.5, + "valid_targets_min": 498 + }, + { + "epoch": 5.2453142227122385, + "grad_norm": 0.9081659373124049, + "learning_rate": 7.184537820068194e-06, + "loss": 0.112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045568905770778656, + "step": 9515, + "valid_targets_mean": 1020.4, + "valid_targets_min": 658 + }, + { + "epoch": 5.248070562293274, + "grad_norm": 0.8665761338498822, + "learning_rate": 7.1634447885037705e-06, + "loss": 0.1046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054460395127534866, + "step": 9520, + "valid_targets_mean": 1393.4, + "valid_targets_min": 594 + }, + { + "epoch": 5.250826901874311, + "grad_norm": 0.8781291197274624, + "learning_rate": 7.142376008948049e-06, + "loss": 0.1189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0651271641254425, + "step": 9525, + "valid_targets_mean": 1613.1, + "valid_targets_min": 674 + }, + { + "epoch": 5.253583241455347, + "grad_norm": 0.9154715371802895, + "learning_rate": 7.121331521206114e-06, + "loss": 0.1143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044558849185705185, + "step": 9530, + "valid_targets_mean": 1025.6, + "valid_targets_min": 648 + }, + { + "epoch": 5.256339581036384, + "grad_norm": 0.8731908004737827, + "learning_rate": 7.1003113650371645e-06, + "loss": 0.1094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.062044765800237656, + "step": 9535, + "valid_targets_mean": 1291.6, + "valid_targets_min": 545 + }, + { + "epoch": 5.25909592061742, + "grad_norm": 0.8617270324535388, + "learning_rate": 7.079315580154425e-06, + "loss": 0.109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060271844267845154, + "step": 9540, + "valid_targets_mean": 1272.5, + "valid_targets_min": 736 + }, + { + "epoch": 5.261852260198457, + "grad_norm": 0.8550049976366203, + "learning_rate": 7.058344206225074e-06, + "loss": 0.1181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06495463848114014, + "step": 9545, + "valid_targets_mean": 1435.8, + "valid_targets_min": 547 + }, + { + "epoch": 5.264608599779493, + "grad_norm": 0.8428941996489469, + "learning_rate": 7.037397282870171e-06, + "loss": 0.1139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050050508230924606, + "step": 9550, + "valid_targets_mean": 1166.6, + "valid_targets_min": 737 + }, + { + "epoch": 5.2673649393605295, + "grad_norm": 0.8820701377060469, + "learning_rate": 7.016474849664592e-06, + "loss": 0.1163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06357172131538391, + "step": 9555, + "valid_targets_mean": 1414.2, + "valid_targets_min": 764 + }, + { + "epoch": 5.2701212789415655, + "grad_norm": 0.8835490351097003, + "learning_rate": 6.995576946136928e-06, + "loss": 0.1227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06415827572345734, + "step": 9560, + "valid_targets_mean": 1457.8, + "valid_targets_min": 662 + }, + { + "epoch": 5.272877618522602, + "grad_norm": 0.8886082491737777, + "learning_rate": 6.974703611769438e-06, + "loss": 0.1249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056758977472782135, + "step": 9565, + "valid_targets_mean": 1344.4, + "valid_targets_min": 680 + }, + { + "epoch": 5.275633958103638, + "grad_norm": 0.8825337634873373, + "learning_rate": 6.9538548859979574e-06, + "loss": 0.1117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049787767231464386, + "step": 9570, + "valid_targets_mean": 1200.9, + "valid_targets_min": 641 + }, + { + "epoch": 5.278390297684675, + "grad_norm": 0.8089449614756246, + "learning_rate": 6.933030808211825e-06, + "loss": 0.111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05772820860147476, + "step": 9575, + "valid_targets_mean": 1432.2, + "valid_targets_min": 730 + }, + { + "epoch": 5.281146637265711, + "grad_norm": 0.9784091952745256, + "learning_rate": 6.912231417753818e-06, + "loss": 0.1197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06050388514995575, + "step": 9580, + "valid_targets_mean": 1447.6, + "valid_targets_min": 598 + }, + { + "epoch": 5.283902976846748, + "grad_norm": 0.8198805935881315, + "learning_rate": 6.891456753920076e-06, + "loss": 0.1057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05008802190423012, + "step": 9585, + "valid_targets_mean": 1212.6, + "valid_targets_min": 583 + }, + { + "epoch": 5.286659316427784, + "grad_norm": 0.8586851974368411, + "learning_rate": 6.8707068559600095e-06, + "loss": 0.1111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046898193657398224, + "step": 9590, + "valid_targets_mean": 1051.5, + "valid_targets_min": 605 + }, + { + "epoch": 5.289415656008821, + "grad_norm": 0.9125183943261566, + "learning_rate": 6.849981763076255e-06, + "loss": 0.115, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07381565123796463, + "step": 9595, + "valid_targets_mean": 1602.2, + "valid_targets_min": 613 + }, + { + "epoch": 5.2921719955898565, + "grad_norm": 0.9703135249405469, + "learning_rate": 6.829281514424564e-06, + "loss": 0.1135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08136849105358124, + "step": 9600, + "valid_targets_mean": 1667.2, + "valid_targets_min": 578 + }, + { + "epoch": 5.294928335170893, + "grad_norm": 0.857121576923735, + "learning_rate": 6.808606149113779e-06, + "loss": 0.1171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06619464606046677, + "step": 9605, + "valid_targets_mean": 1637.0, + "valid_targets_min": 497 + }, + { + "epoch": 5.297684674751929, + "grad_norm": 0.860233371002247, + "learning_rate": 6.787955706205709e-06, + "loss": 0.1094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06652513146400452, + "step": 9610, + "valid_targets_mean": 1513.6, + "valid_targets_min": 603 + }, + { + "epoch": 5.300441014332966, + "grad_norm": 0.8640181350726488, + "learning_rate": 6.76733022471508e-06, + "loss": 0.1138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06567984819412231, + "step": 9615, + "valid_targets_mean": 1785.1, + "valid_targets_min": 520 + }, + { + "epoch": 5.303197353914002, + "grad_norm": 0.8671968818572542, + "learning_rate": 6.746729743609466e-06, + "loss": 0.1145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05760756507515907, + "step": 9620, + "valid_targets_mean": 1283.2, + "valid_targets_min": 593 + }, + { + "epoch": 5.305953693495039, + "grad_norm": 0.8334109287294955, + "learning_rate": 6.726154301809198e-06, + "loss": 0.1177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05939801037311554, + "step": 9625, + "valid_targets_mean": 1396.6, + "valid_targets_min": 689 + }, + { + "epoch": 5.308710033076075, + "grad_norm": 0.8248847019441059, + "learning_rate": 6.705603938187308e-06, + "loss": 0.1147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.066176638007164, + "step": 9630, + "valid_targets_mean": 1385.5, + "valid_targets_min": 577 + }, + { + "epoch": 5.311466372657112, + "grad_norm": 0.8508057427370452, + "learning_rate": 6.685078691569446e-06, + "loss": 0.1123, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05497894063591957, + "step": 9635, + "valid_targets_mean": 1074.1, + "valid_targets_min": 511 + }, + { + "epoch": 5.314222712238148, + "grad_norm": 0.8200820454667774, + "learning_rate": 6.664578600733811e-06, + "loss": 0.113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057640742510557175, + "step": 9640, + "valid_targets_mean": 1384.9, + "valid_targets_min": 777 + }, + { + "epoch": 5.316979051819184, + "grad_norm": 0.7801214083452183, + "learning_rate": 6.64410370441106e-06, + "loss": 0.1225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06455058604478836, + "step": 9645, + "valid_targets_mean": 1467.6, + "valid_targets_min": 1009 + }, + { + "epoch": 5.31973539140022, + "grad_norm": 0.7790819682649669, + "learning_rate": 6.623654041284282e-06, + "loss": 0.1145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052469294518232346, + "step": 9650, + "valid_targets_mean": 1183.2, + "valid_targets_min": 530 + }, + { + "epoch": 5.322491730981257, + "grad_norm": 0.7690094325884088, + "learning_rate": 6.603229649988867e-06, + "loss": 0.1097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046880535781383514, + "step": 9655, + "valid_targets_mean": 1228.0, + "valid_targets_min": 527 + }, + { + "epoch": 5.325248070562293, + "grad_norm": 0.8530166097032374, + "learning_rate": 6.5828305691124615e-06, + "loss": 0.1158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05986800044775009, + "step": 9660, + "valid_targets_mean": 1422.5, + "valid_targets_min": 1039 + }, + { + "epoch": 5.32800441014333, + "grad_norm": 0.838497009710377, + "learning_rate": 6.562456837194903e-06, + "loss": 0.1192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05262543261051178, + "step": 9665, + "valid_targets_mean": 1187.2, + "valid_targets_min": 601 + }, + { + "epoch": 5.330760749724366, + "grad_norm": 0.8355735405531499, + "learning_rate": 6.54210849272813e-06, + "loss": 0.1168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06809254735708237, + "step": 9670, + "valid_targets_mean": 1574.2, + "valid_targets_min": 589 + }, + { + "epoch": 5.333517089305403, + "grad_norm": 0.8650748000500469, + "learning_rate": 6.521785574156119e-06, + "loss": 0.1156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06422356516122818, + "step": 9675, + "valid_targets_mean": 1395.0, + "valid_targets_min": 659 + }, + { + "epoch": 5.336273428886439, + "grad_norm": 0.9879875402290944, + "learning_rate": 6.501488119874811e-06, + "loss": 0.1178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05837864428758621, + "step": 9680, + "valid_targets_mean": 1179.8, + "valid_targets_min": 521 + }, + { + "epoch": 5.3390297684674755, + "grad_norm": 0.9305557552381095, + "learning_rate": 6.481216168232034e-06, + "loss": 0.1098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034623391926288605, + "step": 9685, + "valid_targets_mean": 873.8, + "valid_targets_min": 485 + }, + { + "epoch": 5.341786108048511, + "grad_norm": 0.93112198866101, + "learning_rate": 6.460969757527438e-06, + "loss": 0.1214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051505640149116516, + "step": 9690, + "valid_targets_mean": 1309.2, + "valid_targets_min": 567 + }, + { + "epoch": 5.344542447629548, + "grad_norm": 0.9496533818677787, + "learning_rate": 6.440748926012408e-06, + "loss": 0.1274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07463517785072327, + "step": 9695, + "valid_targets_mean": 1689.1, + "valid_targets_min": 558 + }, + { + "epoch": 5.347298787210584, + "grad_norm": 0.9036860988784097, + "learning_rate": 6.420553711890025e-06, + "loss": 0.1164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05685115605592728, + "step": 9700, + "valid_targets_mean": 1177.2, + "valid_targets_min": 462 + }, + { + "epoch": 5.350055126791621, + "grad_norm": 0.8254665444403249, + "learning_rate": 6.400384153314953e-06, + "loss": 0.1198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05113176256418228, + "step": 9705, + "valid_targets_mean": 1081.6, + "valid_targets_min": 526 + }, + { + "epoch": 5.352811466372657, + "grad_norm": 0.8541971357466949, + "learning_rate": 6.380240288393389e-06, + "loss": 0.1106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06093474477529526, + "step": 9710, + "valid_targets_mean": 1399.0, + "valid_targets_min": 730 + }, + { + "epoch": 5.355567805953694, + "grad_norm": 0.8527335355090475, + "learning_rate": 6.3601221551829865e-06, + "loss": 0.1134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0580768808722496, + "step": 9715, + "valid_targets_mean": 1232.2, + "valid_targets_min": 591 + }, + { + "epoch": 5.35832414553473, + "grad_norm": 0.8857397059366177, + "learning_rate": 6.340029791692788e-06, + "loss": 0.1154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04917246848344803, + "step": 9720, + "valid_targets_mean": 1135.2, + "valid_targets_min": 649 + }, + { + "epoch": 5.3610804851157665, + "grad_norm": 0.9246422030693986, + "learning_rate": 6.319963235883142e-06, + "loss": 0.1117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038003768771886826, + "step": 9725, + "valid_targets_mean": 886.4, + "valid_targets_min": 484 + }, + { + "epoch": 5.3638368246968025, + "grad_norm": 0.9414891023589012, + "learning_rate": 6.299922525665649e-06, + "loss": 0.1137, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07522699981927872, + "step": 9730, + "valid_targets_mean": 1683.6, + "valid_targets_min": 1143 + }, + { + "epoch": 5.366593164277839, + "grad_norm": 0.9613362201396607, + "learning_rate": 6.279907698903071e-06, + "loss": 0.125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06852707266807556, + "step": 9735, + "valid_targets_mean": 1251.6, + "valid_targets_min": 580 + }, + { + "epoch": 5.369349503858875, + "grad_norm": 0.8428102896642, + "learning_rate": 6.259918793409276e-06, + "loss": 0.1165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05434078350663185, + "step": 9740, + "valid_targets_mean": 1225.8, + "valid_targets_min": 698 + }, + { + "epoch": 5.372105843439912, + "grad_norm": 0.7575282703172949, + "learning_rate": 6.239955846949144e-06, + "loss": 0.1157, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05792708694934845, + "step": 9745, + "valid_targets_mean": 1476.8, + "valid_targets_min": 995 + }, + { + "epoch": 5.374862183020948, + "grad_norm": 0.6243548812729134, + "learning_rate": 6.220018897238538e-06, + "loss": 0.1118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051112256944179535, + "step": 9750, + "valid_targets_mean": 1516.2, + "valid_targets_min": 724 + }, + { + "epoch": 5.377618522601985, + "grad_norm": 0.7836856033604848, + "learning_rate": 6.200107981944181e-06, + "loss": 0.1037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05465930700302124, + "step": 9755, + "valid_targets_mean": 1386.0, + "valid_targets_min": 663 + }, + { + "epoch": 5.380374862183021, + "grad_norm": 0.8517552774798791, + "learning_rate": 6.180223138683619e-06, + "loss": 0.1118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05693493038415909, + "step": 9760, + "valid_targets_mean": 1444.8, + "valid_targets_min": 1123 + }, + { + "epoch": 5.383131201764058, + "grad_norm": 0.9754923761460857, + "learning_rate": 6.160364405025141e-06, + "loss": 0.1114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05610571429133415, + "step": 9765, + "valid_targets_mean": 1276.6, + "valid_targets_min": 613 + }, + { + "epoch": 5.3858875413450935, + "grad_norm": 0.9391631213264191, + "learning_rate": 6.1405318184877e-06, + "loss": 0.1165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060184676200151443, + "step": 9770, + "valid_targets_mean": 1328.6, + "valid_targets_min": 561 + }, + { + "epoch": 5.38864388092613, + "grad_norm": 0.9487005481949836, + "learning_rate": 6.120725416540861e-06, + "loss": 0.1189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0564090758562088, + "step": 9775, + "valid_targets_mean": 1281.2, + "valid_targets_min": 812 + }, + { + "epoch": 5.391400220507166, + "grad_norm": 0.8466864824580511, + "learning_rate": 6.1009452366047075e-06, + "loss": 0.1088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04855794459581375, + "step": 9780, + "valid_targets_mean": 1061.0, + "valid_targets_min": 493 + }, + { + "epoch": 5.394156560088203, + "grad_norm": 0.8945280019986567, + "learning_rate": 6.08119131604979e-06, + "loss": 0.1132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046177517622709274, + "step": 9785, + "valid_targets_mean": 999.1, + "valid_targets_min": 502 + }, + { + "epoch": 5.396912899669239, + "grad_norm": 1.171480164423603, + "learning_rate": 6.061463692197036e-06, + "loss": 0.1803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11787073314189911, + "step": 9790, + "valid_targets_mean": 4841.0, + "valid_targets_min": 1966 + }, + { + "epoch": 5.399669239250276, + "grad_norm": 0.9416813065392459, + "learning_rate": 6.0417624023177145e-06, + "loss": 0.2223, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11586426198482513, + "step": 9795, + "valid_targets_mean": 4936.8, + "valid_targets_min": 1399 + }, + { + "epoch": 5.402425578831312, + "grad_norm": 0.7481278449567395, + "learning_rate": 6.022087483633319e-06, + "loss": 0.2233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13795939087867737, + "step": 9800, + "valid_targets_mean": 5672.8, + "valid_targets_min": 2678 + }, + { + "epoch": 5.405181918412349, + "grad_norm": 0.610287380787619, + "learning_rate": 6.0024389733155275e-06, + "loss": 0.2031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08556321263313293, + "step": 9805, + "valid_targets_mean": 4044.2, + "valid_targets_min": 2322 + }, + { + "epoch": 5.407938257993385, + "grad_norm": 0.606570971697229, + "learning_rate": 5.982816908486122e-06, + "loss": 0.2064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08641669154167175, + "step": 9810, + "valid_targets_mean": 3605.0, + "valid_targets_min": 691 + }, + { + "epoch": 5.410694597574421, + "grad_norm": 0.5422831992773844, + "learning_rate": 5.963221326216931e-06, + "loss": 0.2164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08532252162694931, + "step": 9815, + "valid_targets_mean": 4230.2, + "valid_targets_min": 921 + }, + { + "epoch": 5.413450937155457, + "grad_norm": 0.5425047269273212, + "learning_rate": 5.9436522635297375e-06, + "loss": 0.1977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1044442355632782, + "step": 9820, + "valid_targets_mean": 4725.8, + "valid_targets_min": 2127 + }, + { + "epoch": 5.416207276736494, + "grad_norm": 0.539751518859283, + "learning_rate": 5.924109757396228e-06, + "loss": 0.2125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07553604245185852, + "step": 9825, + "valid_targets_mean": 2875.8, + "valid_targets_min": 775 + }, + { + "epoch": 5.41896361631753, + "grad_norm": 0.47944245137137875, + "learning_rate": 5.904593844737916e-06, + "loss": 0.2029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09801404923200607, + "step": 9830, + "valid_targets_mean": 5394.6, + "valid_targets_min": 3997 + }, + { + "epoch": 5.421719955898567, + "grad_norm": 0.5927221944545833, + "learning_rate": 5.885104562426067e-06, + "loss": 0.2178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09888381510972977, + "step": 9835, + "valid_targets_mean": 3801.0, + "valid_targets_min": 954 + }, + { + "epoch": 5.424476295479603, + "grad_norm": 0.4767517799701314, + "learning_rate": 5.865641947281638e-06, + "loss": 0.2067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10180720686912537, + "step": 9840, + "valid_targets_mean": 5662.0, + "valid_targets_min": 717 + }, + { + "epoch": 5.42723263506064, + "grad_norm": 0.5559992736016816, + "learning_rate": 5.84620603607521e-06, + "loss": 0.2038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11537862569093704, + "step": 9845, + "valid_targets_mean": 4817.2, + "valid_targets_min": 1109 + }, + { + "epoch": 5.429988974641676, + "grad_norm": 0.5217874411035978, + "learning_rate": 5.8267968655269026e-06, + "loss": 0.203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09186949580907822, + "step": 9850, + "valid_targets_mean": 5104.2, + "valid_targets_min": 1374 + }, + { + "epoch": 5.4327453142227125, + "grad_norm": 0.5537373453164333, + "learning_rate": 5.807414472306319e-06, + "loss": 0.2052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08641749620437622, + "step": 9855, + "valid_targets_mean": 3611.5, + "valid_targets_min": 1755 + }, + { + "epoch": 5.435501653803748, + "grad_norm": 0.7105611358824806, + "learning_rate": 5.788058893032469e-06, + "loss": 0.1996, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08727101236581802, + "step": 9860, + "valid_targets_mean": 2758.9, + "valid_targets_min": 863 + }, + { + "epoch": 5.438257993384785, + "grad_norm": 0.6127469087197895, + "learning_rate": 5.7687301642737035e-06, + "loss": 0.1959, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09921777993440628, + "step": 9865, + "valid_targets_mean": 4109.1, + "valid_targets_min": 1229 + }, + { + "epoch": 5.441014332965821, + "grad_norm": 0.6260359429883405, + "learning_rate": 5.749428322547652e-06, + "loss": 0.1943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10844361037015915, + "step": 9870, + "valid_targets_mean": 4061.4, + "valid_targets_min": 2126 + }, + { + "epoch": 5.443770672546858, + "grad_norm": 0.6450954266107578, + "learning_rate": 5.730153404321137e-06, + "loss": 0.1981, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10673345625400543, + "step": 9875, + "valid_targets_mean": 3655.9, + "valid_targets_min": 2053 + }, + { + "epoch": 5.446527012127894, + "grad_norm": 0.6621495512229733, + "learning_rate": 5.710905446010122e-06, + "loss": 0.2027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10499429702758789, + "step": 9880, + "valid_targets_mean": 3319.6, + "valid_targets_min": 630 + }, + { + "epoch": 5.449283351708931, + "grad_norm": 0.5694895848363754, + "learning_rate": 5.691684483979629e-06, + "loss": 0.1987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07825897634029388, + "step": 9885, + "valid_targets_mean": 3465.5, + "valid_targets_min": 1581 + }, + { + "epoch": 5.452039691289967, + "grad_norm": 0.6190833859392262, + "learning_rate": 5.672490554543673e-06, + "loss": 0.2036, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10048248618841171, + "step": 9890, + "valid_targets_mean": 3570.1, + "valid_targets_min": 960 + }, + { + "epoch": 5.4547960308710035, + "grad_norm": 0.6239425582721292, + "learning_rate": 5.653323693965218e-06, + "loss": 0.1847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11257928609848022, + "step": 9895, + "valid_targets_mean": 4040.0, + "valid_targets_min": 2297 + }, + { + "epoch": 5.4575523704520394, + "grad_norm": 0.5965570637007923, + "learning_rate": 5.634183938456061e-06, + "loss": 0.1876, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.099518783390522, + "step": 9900, + "valid_targets_mean": 4187.5, + "valid_targets_min": 1749 + }, + { + "epoch": 5.460308710033076, + "grad_norm": 0.7089165902186726, + "learning_rate": 5.615071324176806e-06, + "loss": 0.1845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12451255321502686, + "step": 9905, + "valid_targets_mean": 4211.1, + "valid_targets_min": 2076 + }, + { + "epoch": 5.463065049614112, + "grad_norm": 0.6448718037440176, + "learning_rate": 5.59598588723677e-06, + "loss": 0.1986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0966302752494812, + "step": 9910, + "valid_targets_mean": 3433.4, + "valid_targets_min": 1743 + }, + { + "epoch": 5.465821389195149, + "grad_norm": 0.6260969345166869, + "learning_rate": 5.576927663693929e-06, + "loss": 0.197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08555245399475098, + "step": 9915, + "valid_targets_mean": 3017.8, + "valid_targets_min": 1043 + }, + { + "epoch": 5.468577728776185, + "grad_norm": 0.6471130593266455, + "learning_rate": 5.557896689554843e-06, + "loss": 0.1852, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08888323605060577, + "step": 9920, + "valid_targets_mean": 4154.0, + "valid_targets_min": 956 + }, + { + "epoch": 5.471334068357222, + "grad_norm": 0.9406983595997962, + "learning_rate": 5.53889300077459e-06, + "loss": 0.1872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08435867726802826, + "step": 9925, + "valid_targets_mean": 4138.5, + "valid_targets_min": 1389 + }, + { + "epoch": 5.474090407938258, + "grad_norm": 0.6075652228220261, + "learning_rate": 5.519916633256699e-06, + "loss": 0.1905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08267098665237427, + "step": 9930, + "valid_targets_mean": 3197.0, + "valid_targets_min": 1997 + }, + { + "epoch": 5.476846747519295, + "grad_norm": 0.6367959761902716, + "learning_rate": 5.500967622853073e-06, + "loss": 0.1905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09354935586452484, + "step": 9935, + "valid_targets_mean": 3766.9, + "valid_targets_min": 2132 + }, + { + "epoch": 5.4796030871003305, + "grad_norm": 0.5976961203211802, + "learning_rate": 5.482046005363952e-06, + "loss": 0.1819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08770637959241867, + "step": 9940, + "valid_targets_mean": 3891.9, + "valid_targets_min": 768 + }, + { + "epoch": 5.482359426681367, + "grad_norm": 0.6620466890858857, + "learning_rate": 5.463151816537802e-06, + "loss": 0.1883, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07340322434902191, + "step": 9945, + "valid_targets_mean": 2598.5, + "valid_targets_min": 940 + }, + { + "epoch": 5.485115766262403, + "grad_norm": 0.7060374473071853, + "learning_rate": 5.4442850920712704e-06, + "loss": 0.1875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0769575983285904, + "step": 9950, + "valid_targets_mean": 2420.8, + "valid_targets_min": 1118 + }, + { + "epoch": 5.48787210584344, + "grad_norm": 0.6032642227474491, + "learning_rate": 5.4254458676091245e-06, + "loss": 0.178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07656289637088776, + "step": 9955, + "valid_targets_mean": 3099.2, + "valid_targets_min": 1541 + }, + { + "epoch": 5.490628445424476, + "grad_norm": 0.7691161200884521, + "learning_rate": 5.406634178744168e-06, + "loss": 0.1896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11048603057861328, + "step": 9960, + "valid_targets_mean": 2600.6, + "valid_targets_min": 617 + }, + { + "epoch": 5.493384785005513, + "grad_norm": 0.8020976983824479, + "learning_rate": 5.387850061017186e-06, + "loss": 0.1874, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10676752030849457, + "step": 9965, + "valid_targets_mean": 2638.8, + "valid_targets_min": 888 + }, + { + "epoch": 5.496141124586549, + "grad_norm": 0.671107730104951, + "learning_rate": 5.369093549916873e-06, + "loss": 0.1876, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09959806501865387, + "step": 9970, + "valid_targets_mean": 3712.1, + "valid_targets_min": 2268 + }, + { + "epoch": 5.498897464167586, + "grad_norm": 0.6727201958731953, + "learning_rate": 5.350364680879769e-06, + "loss": 0.1858, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08801421523094177, + "step": 9975, + "valid_targets_mean": 2838.0, + "valid_targets_min": 1840 + }, + { + "epoch": 5.501653803748622, + "grad_norm": 0.6851718594846113, + "learning_rate": 5.331663489290184e-06, + "loss": 0.1844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07932449877262115, + "step": 9980, + "valid_targets_mean": 2626.9, + "valid_targets_min": 761 + }, + { + "epoch": 5.504410143329658, + "grad_norm": 0.6615606013056048, + "learning_rate": 5.312990010480137e-06, + "loss": 0.189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0840315967798233, + "step": 9985, + "valid_targets_mean": 2880.5, + "valid_targets_min": 826 + }, + { + "epoch": 5.507166482910694, + "grad_norm": 0.6386126009379645, + "learning_rate": 5.294344279729307e-06, + "loss": 0.1879, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08127246797084808, + "step": 9990, + "valid_targets_mean": 3267.4, + "valid_targets_min": 1727 + }, + { + "epoch": 5.509922822491731, + "grad_norm": 0.6585724540888849, + "learning_rate": 5.2757263322649235e-06, + "loss": 0.1853, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07358944416046143, + "step": 9995, + "valid_targets_mean": 3612.2, + "valid_targets_min": 989 + }, + { + "epoch": 5.512679162072767, + "grad_norm": 0.6396847818888965, + "learning_rate": 5.257136203261741e-06, + "loss": 0.1921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10030101239681244, + "step": 10000, + "valid_targets_mean": 3919.2, + "valid_targets_min": 1076 + }, + { + "epoch": 5.515435501653804, + "grad_norm": 0.6437937890477636, + "learning_rate": 5.2385739278419525e-06, + "loss": 0.1819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10403428226709366, + "step": 10005, + "valid_targets_mean": 4062.1, + "valid_targets_min": 1537 + }, + { + "epoch": 5.51819184123484, + "grad_norm": 0.6362044472916596, + "learning_rate": 5.220039541075126e-06, + "loss": 0.1846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10561394691467285, + "step": 10010, + "valid_targets_mean": 4059.4, + "valid_targets_min": 2065 + }, + { + "epoch": 5.520948180815877, + "grad_norm": 0.6962679897344297, + "learning_rate": 5.201533077978141e-06, + "loss": 0.1834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09952652454376221, + "step": 10015, + "valid_targets_mean": 3896.6, + "valid_targets_min": 2047 + }, + { + "epoch": 5.523704520396913, + "grad_norm": 0.6412612064295296, + "learning_rate": 5.18305457351512e-06, + "loss": 0.1889, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10712067037820816, + "step": 10020, + "valid_targets_mean": 3880.2, + "valid_targets_min": 1307 + }, + { + "epoch": 5.5264608599779494, + "grad_norm": 0.7205798231007587, + "learning_rate": 5.164604062597369e-06, + "loss": 0.1892, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0939706340432167, + "step": 10025, + "valid_targets_mean": 2856.6, + "valid_targets_min": 807 + }, + { + "epoch": 5.529217199558985, + "grad_norm": 0.6301843549788713, + "learning_rate": 5.1461815800832935e-06, + "loss": 0.1925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07916659116744995, + "step": 10030, + "valid_targets_mean": 2880.9, + "valid_targets_min": 1201 + }, + { + "epoch": 5.531973539140022, + "grad_norm": 0.610033427340723, + "learning_rate": 5.127787160778366e-06, + "loss": 0.1845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07521520555019379, + "step": 10035, + "valid_targets_mean": 3004.5, + "valid_targets_min": 300 + }, + { + "epoch": 5.534729878721058, + "grad_norm": 0.6274034140855116, + "learning_rate": 5.109420839435018e-06, + "loss": 0.1876, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09258857369422913, + "step": 10040, + "valid_targets_mean": 3600.9, + "valid_targets_min": 2462 + }, + { + "epoch": 5.537486218302095, + "grad_norm": 0.6357837103591361, + "learning_rate": 5.0910826507526055e-06, + "loss": 0.1891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0989026129245758, + "step": 10045, + "valid_targets_mean": 4155.0, + "valid_targets_min": 1227 + }, + { + "epoch": 5.540242557883131, + "grad_norm": 0.6546711244977618, + "learning_rate": 5.072772629377334e-06, + "loss": 0.185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0818139910697937, + "step": 10050, + "valid_targets_mean": 3017.0, + "valid_targets_min": 964 + }, + { + "epoch": 5.542998897464168, + "grad_norm": 0.6086784289344047, + "learning_rate": 5.054490809902188e-06, + "loss": 0.1761, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08657610416412354, + "step": 10055, + "valid_targets_mean": 3252.9, + "valid_targets_min": 2045 + }, + { + "epoch": 5.545755237045204, + "grad_norm": 0.6634804292485227, + "learning_rate": 5.03623722686688e-06, + "loss": 0.1845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09933739900588989, + "step": 10060, + "valid_targets_mean": 4026.8, + "valid_targets_min": 840 + }, + { + "epoch": 5.5485115766262405, + "grad_norm": 0.640561858561461, + "learning_rate": 5.018011914757763e-06, + "loss": 0.1709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09852005541324615, + "step": 10065, + "valid_targets_mean": 4145.4, + "valid_targets_min": 512 + }, + { + "epoch": 5.551267916207276, + "grad_norm": 0.644749189548376, + "learning_rate": 4.999814908007787e-06, + "loss": 0.2009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10495756566524506, + "step": 10070, + "valid_targets_mean": 3610.0, + "valid_targets_min": 1651 + }, + { + "epoch": 5.554024255788313, + "grad_norm": 0.6544137723789439, + "learning_rate": 4.981646240996421e-06, + "loss": 0.1859, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07770076394081116, + "step": 10075, + "valid_targets_mean": 2796.1, + "valid_targets_min": 1739 + }, + { + "epoch": 5.556780595369349, + "grad_norm": 0.6232035382256774, + "learning_rate": 4.963505948049587e-06, + "loss": 0.1911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0872005820274353, + "step": 10080, + "valid_targets_mean": 3076.9, + "valid_targets_min": 594 + }, + { + "epoch": 5.559536934950386, + "grad_norm": 0.6405652800294387, + "learning_rate": 4.945394063439621e-06, + "loss": 0.1901, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09569260478019714, + "step": 10085, + "valid_targets_mean": 3337.9, + "valid_targets_min": 1566 + }, + { + "epoch": 5.562293274531422, + "grad_norm": 0.7092131170422192, + "learning_rate": 4.927310621385162e-06, + "loss": 0.1919, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09721282869577408, + "step": 10090, + "valid_targets_mean": 3851.8, + "valid_targets_min": 1213 + }, + { + "epoch": 5.565049614112459, + "grad_norm": 0.672414160027086, + "learning_rate": 4.909255656051126e-06, + "loss": 0.1862, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09909454733133316, + "step": 10095, + "valid_targets_mean": 2973.0, + "valid_targets_min": 1215 + }, + { + "epoch": 5.567805953693495, + "grad_norm": 0.6483131677032945, + "learning_rate": 4.891229201548626e-06, + "loss": 0.1729, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09791943430900574, + "step": 10100, + "valid_targets_mean": 4297.1, + "valid_targets_min": 2716 + }, + { + "epoch": 5.570562293274532, + "grad_norm": 0.6069493442619707, + "learning_rate": 4.873231291934905e-06, + "loss": 0.1765, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09416918456554413, + "step": 10105, + "valid_targets_mean": 3919.9, + "valid_targets_min": 1955 + }, + { + "epoch": 5.5733186328555675, + "grad_norm": 0.7031182206365024, + "learning_rate": 4.855261961213285e-06, + "loss": 0.1738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09127628803253174, + "step": 10110, + "valid_targets_mean": 3298.9, + "valid_targets_min": 643 + }, + { + "epoch": 5.576074972436604, + "grad_norm": 0.682649392407533, + "learning_rate": 4.8373212433330884e-06, + "loss": 0.1741, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08964389562606812, + "step": 10115, + "valid_targets_mean": 3446.2, + "valid_targets_min": 966 + }, + { + "epoch": 5.57883131201764, + "grad_norm": 0.641961897211857, + "learning_rate": 4.819409172189578e-06, + "loss": 0.1703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07452800869941711, + "step": 10120, + "valid_targets_mean": 2767.8, + "valid_targets_min": 1101 + }, + { + "epoch": 5.581587651598677, + "grad_norm": 0.6836693595638798, + "learning_rate": 4.801525781623899e-06, + "loss": 0.1662, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08057554066181183, + "step": 10125, + "valid_targets_mean": 2842.2, + "valid_targets_min": 610 + }, + { + "epoch": 5.584343991179713, + "grad_norm": 0.7270236945961366, + "learning_rate": 4.7836711054230025e-06, + "loss": 0.1811, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10653100162744522, + "step": 10130, + "valid_targets_mean": 3791.8, + "valid_targets_min": 1285 + }, + { + "epoch": 5.58710033076075, + "grad_norm": 0.677918888125832, + "learning_rate": 4.765845177319608e-06, + "loss": 0.1986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08355588465929031, + "step": 10135, + "valid_targets_mean": 3571.8, + "valid_targets_min": 1236 + }, + { + "epoch": 5.589856670341786, + "grad_norm": 0.6702513647383587, + "learning_rate": 4.748048030992105e-06, + "loss": 0.1755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10104469954967499, + "step": 10140, + "valid_targets_mean": 3953.5, + "valid_targets_min": 2520 + }, + { + "epoch": 5.592613009922823, + "grad_norm": 0.6818321412764785, + "learning_rate": 4.730279700064508e-06, + "loss": 0.1837, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09449208527803421, + "step": 10145, + "valid_targets_mean": 3936.1, + "valid_targets_min": 1136 + }, + { + "epoch": 5.595369349503859, + "grad_norm": 0.6785122155847518, + "learning_rate": 4.712540218106394e-06, + "loss": 0.1734, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09206905961036682, + "step": 10150, + "valid_targets_mean": 3288.1, + "valid_targets_min": 1728 + }, + { + "epoch": 5.598125689084895, + "grad_norm": 0.6444729366758138, + "learning_rate": 4.694829618632834e-06, + "loss": 0.183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07870101183652878, + "step": 10155, + "valid_targets_mean": 3756.9, + "valid_targets_min": 1631 + }, + { + "epoch": 5.600882028665931, + "grad_norm": 0.6426016312849233, + "learning_rate": 4.677147935104334e-06, + "loss": 0.1794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0771336480975151, + "step": 10160, + "valid_targets_mean": 3669.8, + "valid_targets_min": 2408 + }, + { + "epoch": 5.603638368246968, + "grad_norm": 0.6128905139829464, + "learning_rate": 4.659495200926767e-06, + "loss": 0.1684, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09397043287754059, + "step": 10165, + "valid_targets_mean": 3905.9, + "valid_targets_min": 1763 + }, + { + "epoch": 5.606394707828004, + "grad_norm": 0.9897675756442681, + "learning_rate": 4.6418714494513115e-06, + "loss": 0.1895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10874296724796295, + "step": 10170, + "valid_targets_mean": 2997.2, + "valid_targets_min": 1643 + }, + { + "epoch": 5.609151047409041, + "grad_norm": 0.7512517317548107, + "learning_rate": 4.624276713974389e-06, + "loss": 0.1716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07776353508234024, + "step": 10175, + "valid_targets_mean": 2774.0, + "valid_targets_min": 1273 + }, + { + "epoch": 5.611907386990077, + "grad_norm": 0.7091076580919284, + "learning_rate": 4.60671102773761e-06, + "loss": 0.1807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08871421962976456, + "step": 10180, + "valid_targets_mean": 2949.0, + "valid_targets_min": 770 + }, + { + "epoch": 5.614663726571114, + "grad_norm": 0.6751556974759574, + "learning_rate": 4.589174423927692e-06, + "loss": 0.1845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07350783050060272, + "step": 10185, + "valid_targets_mean": 2409.8, + "valid_targets_min": 364 + }, + { + "epoch": 5.61742006615215, + "grad_norm": 0.6228243074046986, + "learning_rate": 4.571666935676411e-06, + "loss": 0.1835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09622617065906525, + "step": 10190, + "valid_targets_mean": 4212.2, + "valid_targets_min": 1168 + }, + { + "epoch": 5.620176405733186, + "grad_norm": 0.7464174253219379, + "learning_rate": 4.554188596060532e-06, + "loss": 0.1963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11194835603237152, + "step": 10195, + "valid_targets_mean": 3440.5, + "valid_targets_min": 1560 + }, + { + "epoch": 5.622932745314222, + "grad_norm": 0.666392662382764, + "learning_rate": 4.536739438101756e-06, + "loss": 0.1769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08386045694351196, + "step": 10200, + "valid_targets_mean": 2950.2, + "valid_targets_min": 1373 + }, + { + "epoch": 5.625689084895259, + "grad_norm": 0.6227745531283247, + "learning_rate": 4.519319494766645e-06, + "loss": 0.1945, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07452835887670517, + "step": 10205, + "valid_targets_mean": 3080.8, + "valid_targets_min": 1435 + }, + { + "epoch": 5.628445424476295, + "grad_norm": 0.6911787111079792, + "learning_rate": 4.501928798966573e-06, + "loss": 0.1831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11125342547893524, + "step": 10210, + "valid_targets_mean": 4122.6, + "valid_targets_min": 2204 + }, + { + "epoch": 5.631201764057332, + "grad_norm": 0.5898210399993414, + "learning_rate": 4.484567383557648e-06, + "loss": 0.1829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10456360131502151, + "step": 10215, + "valid_targets_mean": 5108.1, + "valid_targets_min": 2223 + }, + { + "epoch": 5.633958103638368, + "grad_norm": 0.6446636750682415, + "learning_rate": 4.467235281340667e-06, + "loss": 0.1877, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10473263263702393, + "step": 10220, + "valid_targets_mean": 4091.5, + "valid_targets_min": 1137 + }, + { + "epoch": 5.636714443219405, + "grad_norm": 0.8502794329998878, + "learning_rate": 4.449932525061036e-06, + "loss": 0.1791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07407791912555695, + "step": 10225, + "valid_targets_mean": 3131.0, + "valid_targets_min": 373 + }, + { + "epoch": 5.639470782800441, + "grad_norm": 0.6588196683723032, + "learning_rate": 4.432659147408736e-06, + "loss": 0.1804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07570677995681763, + "step": 10230, + "valid_targets_mean": 2893.9, + "valid_targets_min": 352 + }, + { + "epoch": 5.6422271223814775, + "grad_norm": 0.6197662728337073, + "learning_rate": 4.415415181018223e-06, + "loss": 0.1846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08128007501363754, + "step": 10235, + "valid_targets_mean": 4443.4, + "valid_targets_min": 2838 + }, + { + "epoch": 5.644983461962513, + "grad_norm": 0.707286040627471, + "learning_rate": 4.398200658468397e-06, + "loss": 0.1746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09532670676708221, + "step": 10240, + "valid_targets_mean": 3518.8, + "valid_targets_min": 1369 + }, + { + "epoch": 5.64773980154355, + "grad_norm": 0.7051676098522763, + "learning_rate": 4.381015612282529e-06, + "loss": 0.1716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10902032256126404, + "step": 10245, + "valid_targets_mean": 4071.6, + "valid_targets_min": 2579 + }, + { + "epoch": 5.650496141124586, + "grad_norm": 0.6604293789009915, + "learning_rate": 4.363860074928197e-06, + "loss": 0.1757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06612522900104523, + "step": 10250, + "valid_targets_mean": 2293.9, + "valid_targets_min": 306 + }, + { + "epoch": 5.653252480705623, + "grad_norm": 0.669528919083904, + "learning_rate": 4.346734078817234e-06, + "loss": 0.1737, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09303723275661469, + "step": 10255, + "valid_targets_mean": 3565.0, + "valid_targets_min": 1399 + }, + { + "epoch": 5.656008820286659, + "grad_norm": 0.6399352352729987, + "learning_rate": 4.329637656305654e-06, + "loss": 0.181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08346434682607651, + "step": 10260, + "valid_targets_mean": 3098.2, + "valid_targets_min": 1062 + }, + { + "epoch": 5.658765159867696, + "grad_norm": 0.7111714944706979, + "learning_rate": 4.312570839693606e-06, + "loss": 0.1784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09731197357177734, + "step": 10265, + "valid_targets_mean": 3322.4, + "valid_targets_min": 1093 + }, + { + "epoch": 5.661521499448732, + "grad_norm": 0.6037401074298192, + "learning_rate": 4.295533661225295e-06, + "loss": 0.1803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08997463434934616, + "step": 10270, + "valid_targets_mean": 3481.9, + "valid_targets_min": 313 + }, + { + "epoch": 5.664277839029769, + "grad_norm": 0.7303847971451407, + "learning_rate": 4.2785261530889335e-06, + "loss": 0.1764, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08168359845876694, + "step": 10275, + "valid_targets_mean": 2449.4, + "valid_targets_min": 783 + }, + { + "epoch": 5.6670341786108045, + "grad_norm": 0.7099252456172216, + "learning_rate": 4.261548347416691e-06, + "loss": 0.1863, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10832295566797256, + "step": 10280, + "valid_targets_mean": 4458.8, + "valid_targets_min": 1177 + }, + { + "epoch": 5.669790518191841, + "grad_norm": 0.6931442767258408, + "learning_rate": 4.244600276284607e-06, + "loss": 0.1777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07646596431732178, + "step": 10285, + "valid_targets_mean": 3365.0, + "valid_targets_min": 1972 + }, + { + "epoch": 5.672546857772877, + "grad_norm": 0.718987774087432, + "learning_rate": 4.227681971712543e-06, + "loss": 0.1669, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09947442263364792, + "step": 10290, + "valid_targets_mean": 4161.5, + "valid_targets_min": 2550 + }, + { + "epoch": 5.675303197353914, + "grad_norm": 0.5993608154736098, + "learning_rate": 4.210793465664127e-06, + "loss": 0.1736, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08648347854614258, + "step": 10295, + "valid_targets_mean": 6125.4, + "valid_targets_min": 2132 + }, + { + "epoch": 5.67805953693495, + "grad_norm": 0.6344882972024648, + "learning_rate": 4.19393479004669e-06, + "loss": 0.1756, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08712658286094666, + "step": 10300, + "valid_targets_mean": 3607.5, + "valid_targets_min": 2280 + }, + { + "epoch": 5.680815876515987, + "grad_norm": 0.6775566416416411, + "learning_rate": 4.177105976711204e-06, + "loss": 0.171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07299946248531342, + "step": 10305, + "valid_targets_mean": 2645.5, + "valid_targets_min": 794 + }, + { + "epoch": 5.683572216097023, + "grad_norm": 0.6371248470258349, + "learning_rate": 4.160307057452218e-06, + "loss": 0.1658, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08123248815536499, + "step": 10310, + "valid_targets_mean": 3561.2, + "valid_targets_min": 1293 + }, + { + "epoch": 5.68632855567806, + "grad_norm": 0.6735817382543742, + "learning_rate": 4.143538064007808e-06, + "loss": 0.168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08925417065620422, + "step": 10315, + "valid_targets_mean": 3339.4, + "valid_targets_min": 1357 + }, + { + "epoch": 5.6890848952590956, + "grad_norm": 0.6521145591183494, + "learning_rate": 4.126799028059505e-06, + "loss": 0.1698, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08456301689147949, + "step": 10320, + "valid_targets_mean": 3513.9, + "valid_targets_min": 2360 + }, + { + "epoch": 5.691841234840132, + "grad_norm": 0.6631660768679856, + "learning_rate": 4.110089981232253e-06, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10367235541343689, + "step": 10325, + "valid_targets_mean": 3954.4, + "valid_targets_min": 1213 + }, + { + "epoch": 5.694597574421168, + "grad_norm": 0.6693259173403967, + "learning_rate": 4.093410955094328e-06, + "loss": 0.1726, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08421348035335541, + "step": 10330, + "valid_targets_mean": 3356.9, + "valid_targets_min": 1708 + }, + { + "epoch": 5.697353914002205, + "grad_norm": 0.750307971836502, + "learning_rate": 4.076761981157289e-06, + "loss": 0.1736, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09388542175292969, + "step": 10335, + "valid_targets_mean": 2848.6, + "valid_targets_min": 1242 + }, + { + "epoch": 5.700110253583241, + "grad_norm": 0.7452090293664275, + "learning_rate": 4.060143090875921e-06, + "loss": 0.1622, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08988501131534576, + "step": 10340, + "valid_targets_mean": 2673.5, + "valid_targets_min": 750 + }, + { + "epoch": 5.702866593164278, + "grad_norm": 0.7520900325445101, + "learning_rate": 4.043554315648166e-06, + "loss": 0.1803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07388913631439209, + "step": 10345, + "valid_targets_mean": 2704.1, + "valid_targets_min": 280 + }, + { + "epoch": 5.705622932745314, + "grad_norm": 0.8040138727964641, + "learning_rate": 4.026995686815081e-06, + "loss": 0.1751, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09325580298900604, + "step": 10350, + "valid_targets_mean": 2230.9, + "valid_targets_min": 313 + }, + { + "epoch": 5.708379272326351, + "grad_norm": 0.6743141064556266, + "learning_rate": 4.010467235660753e-06, + "loss": 0.1822, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0868203416466713, + "step": 10355, + "valid_targets_mean": 3992.8, + "valid_targets_min": 2193 + }, + { + "epoch": 5.711135611907387, + "grad_norm": 0.6888531947622752, + "learning_rate": 3.9939689934122694e-06, + "loss": 0.1747, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09071101248264313, + "step": 10360, + "valid_targets_mean": 3137.5, + "valid_targets_min": 325 + }, + { + "epoch": 5.713891951488423, + "grad_norm": 0.6616452600574442, + "learning_rate": 3.977500991239631e-06, + "loss": 0.1706, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07428516447544098, + "step": 10365, + "valid_targets_mean": 3164.1, + "valid_targets_min": 1286 + }, + { + "epoch": 5.716648291069459, + "grad_norm": 0.6928629462189785, + "learning_rate": 3.961063260255713e-06, + "loss": 0.1868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09862911701202393, + "step": 10370, + "valid_targets_mean": 3985.8, + "valid_targets_min": 1765 + }, + { + "epoch": 5.719404630650496, + "grad_norm": 0.5161501660447919, + "learning_rate": 3.944655831516204e-06, + "loss": 0.1473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03955835476517677, + "step": 10375, + "valid_targets_mean": 2390.1, + "valid_targets_min": 872 + }, + { + "epoch": 5.722160970231532, + "grad_norm": 0.5098691115557363, + "learning_rate": 3.928278736019538e-06, + "loss": 0.0952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03330495208501816, + "step": 10380, + "valid_targets_mean": 1543.9, + "valid_targets_min": 665 + }, + { + "epoch": 5.724917309812569, + "grad_norm": 0.3729845433292964, + "learning_rate": 3.911932004706836e-06, + "loss": 0.0725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04079446941614151, + "step": 10385, + "valid_targets_mean": 3129.8, + "valid_targets_min": 2254 + }, + { + "epoch": 5.727673649393605, + "grad_norm": 0.41017804951102504, + "learning_rate": 3.895615668461856e-06, + "loss": 0.072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03468386083841324, + "step": 10390, + "valid_targets_mean": 3026.8, + "valid_targets_min": 706 + }, + { + "epoch": 5.730429988974642, + "grad_norm": 0.3358441232580079, + "learning_rate": 3.8793297581109325e-06, + "loss": 0.0829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026570318266749382, + "step": 10395, + "valid_targets_mean": 3207.8, + "valid_targets_min": 1216 + }, + { + "epoch": 5.733186328555679, + "grad_norm": 0.39805107384827454, + "learning_rate": 3.863074304422918e-06, + "loss": 0.073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04147864505648613, + "step": 10400, + "valid_targets_mean": 3540.0, + "valid_targets_min": 2551 + }, + { + "epoch": 5.7359426681367145, + "grad_norm": 0.7524776952314952, + "learning_rate": 3.846849338109116e-06, + "loss": 0.1076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052572883665561676, + "step": 10405, + "valid_targets_mean": 960.9, + "valid_targets_min": 555 + }, + { + "epoch": 5.73869900771775, + "grad_norm": 0.3769373594023775, + "learning_rate": 3.830654889823238e-06, + "loss": 0.0998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03400315344333649, + "step": 10410, + "valid_targets_mean": 3465.4, + "valid_targets_min": 1008 + }, + { + "epoch": 5.741455347298787, + "grad_norm": 0.4251243762797158, + "learning_rate": 3.8144909901613303e-06, + "loss": 0.0835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042783044278621674, + "step": 10415, + "valid_targets_mean": 3522.1, + "valid_targets_min": 3039 + }, + { + "epoch": 5.744211686879824, + "grad_norm": 0.4521224988849751, + "learning_rate": 3.7983576696617364e-06, + "loss": 0.1128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06753584742546082, + "step": 10420, + "valid_targets_mean": 3348.0, + "valid_targets_min": 1938 + }, + { + "epoch": 5.74696802646086, + "grad_norm": 0.5948279046067094, + "learning_rate": 3.7822549588050185e-06, + "loss": 0.0629, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037634119391441345, + "step": 10425, + "valid_targets_mean": 3973.6, + "valid_targets_min": 924 + }, + { + "epoch": 5.749724366041896, + "grad_norm": 0.4462453849269703, + "learning_rate": 3.766182888013905e-06, + "loss": 0.0737, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03821098804473877, + "step": 10430, + "valid_targets_mean": 2262.8, + "valid_targets_min": 444 + }, + { + "epoch": 5.752480705622933, + "grad_norm": 0.37896058848035463, + "learning_rate": 3.7501414876532425e-06, + "loss": 0.0738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0366894006729126, + "step": 10435, + "valid_targets_mean": 3283.2, + "valid_targets_min": 749 + }, + { + "epoch": 5.75523704520397, + "grad_norm": 0.5132760394590526, + "learning_rate": 3.7341307880299306e-06, + "loss": 0.0932, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04661383479833603, + "step": 10440, + "valid_targets_mean": 1381.4, + "valid_targets_min": 601 + }, + { + "epoch": 5.7579933847850056, + "grad_norm": 0.4618734947279742, + "learning_rate": 3.718150819392863e-06, + "loss": 0.0895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03807876631617546, + "step": 10445, + "valid_targets_mean": 3375.0, + "valid_targets_min": 1756 + }, + { + "epoch": 5.7607497243660415, + "grad_norm": 0.3854023685738128, + "learning_rate": 3.7022016119328763e-06, + "loss": 0.0717, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02920461632311344, + "step": 10450, + "valid_targets_mean": 2760.6, + "valid_targets_min": 571 + }, + { + "epoch": 5.763506063947078, + "grad_norm": 0.48242363197817867, + "learning_rate": 3.686283195782692e-06, + "loss": 0.0683, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0330260805785656, + "step": 10455, + "valid_targets_mean": 2584.8, + "valid_targets_min": 520 + }, + { + "epoch": 5.766262403528115, + "grad_norm": 0.47997933534452625, + "learning_rate": 3.670395601016854e-06, + "loss": 0.0733, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04837290197610855, + "step": 10460, + "valid_targets_mean": 3460.2, + "valid_targets_min": 2179 + }, + { + "epoch": 5.769018743109151, + "grad_norm": 0.38725706295277573, + "learning_rate": 3.654538857651675e-06, + "loss": 0.0771, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030730318278074265, + "step": 10465, + "valid_targets_mean": 3006.9, + "valid_targets_min": 807 + }, + { + "epoch": 5.771775082690187, + "grad_norm": 0.481120724508489, + "learning_rate": 3.638712995645188e-06, + "loss": 0.0836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03414568677544594, + "step": 10470, + "valid_targets_mean": 1811.0, + "valid_targets_min": 594 + }, + { + "epoch": 5.774531422271224, + "grad_norm": 0.5056209971945618, + "learning_rate": 3.622918044897077e-06, + "loss": 0.0833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04170460253953934, + "step": 10475, + "valid_targets_mean": 2210.0, + "valid_targets_min": 714 + }, + { + "epoch": 5.777287761852261, + "grad_norm": 0.5241364692869734, + "learning_rate": 3.6071540352486234e-06, + "loss": 0.1008, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08811280131340027, + "step": 10480, + "valid_targets_mean": 3098.4, + "valid_targets_min": 1227 + }, + { + "epoch": 5.780044101433297, + "grad_norm": 0.4629595704143088, + "learning_rate": 3.591420996482655e-06, + "loss": 0.0846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03343856334686279, + "step": 10485, + "valid_targets_mean": 2710.6, + "valid_targets_min": 782 + }, + { + "epoch": 5.7828004410143325, + "grad_norm": 0.6574083598554025, + "learning_rate": 3.5757189583234817e-06, + "loss": 0.0742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057368628680706024, + "step": 10490, + "valid_targets_mean": 1954.5, + "valid_targets_min": 950 + }, + { + "epoch": 5.785556780595369, + "grad_norm": 0.42138383230575976, + "learning_rate": 3.5600479504368533e-06, + "loss": 0.1026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04019131511449814, + "step": 10495, + "valid_targets_mean": 3065.0, + "valid_targets_min": 1296 + }, + { + "epoch": 5.788313120176406, + "grad_norm": 0.4122534340228617, + "learning_rate": 3.5444080024298844e-06, + "loss": 0.0911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028338219970464706, + "step": 10500, + "valid_targets_mean": 3146.6, + "valid_targets_min": 789 + }, + { + "epoch": 5.791069459757442, + "grad_norm": 0.5425461254608329, + "learning_rate": 3.528799143851016e-06, + "loss": 0.0784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048115238547325134, + "step": 10505, + "valid_targets_mean": 2301.5, + "valid_targets_min": 713 + }, + { + "epoch": 5.793825799338478, + "grad_norm": 0.5658698811561962, + "learning_rate": 3.5132214041899482e-06, + "loss": 0.2458, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10188327729701996, + "step": 10510, + "valid_targets_mean": 2630.1, + "valid_targets_min": 980 + }, + { + "epoch": 5.796582138919515, + "grad_norm": 0.6241294583977877, + "learning_rate": 3.497674812877583e-06, + "loss": 0.0946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07481149584054947, + "step": 10515, + "valid_targets_mean": 1792.1, + "valid_targets_min": 753 + }, + { + "epoch": 5.799338478500552, + "grad_norm": 0.4584492621336809, + "learning_rate": 3.4821593992859893e-06, + "loss": 0.081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03224394470453262, + "step": 10520, + "valid_targets_mean": 1459.5, + "valid_targets_min": 670 + }, + { + "epoch": 5.802094818081588, + "grad_norm": 0.4726368303109881, + "learning_rate": 3.4666751927283217e-06, + "loss": 0.1676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0435740202665329, + "step": 10525, + "valid_targets_mean": 2546.0, + "valid_targets_min": 663 + }, + { + "epoch": 5.804851157662624, + "grad_norm": 0.3135063123023392, + "learning_rate": 3.4512222224587742e-06, + "loss": 0.0714, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.024659529328346252, + "step": 10530, + "valid_targets_mean": 5186.2, + "valid_targets_min": 3522 + }, + { + "epoch": 5.80760749724366, + "grad_norm": 0.48017210444202546, + "learning_rate": 3.4358005176725317e-06, + "loss": 0.0827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0499797984957695, + "step": 10535, + "valid_targets_mean": 4087.4, + "valid_targets_min": 2327 + }, + { + "epoch": 5.810363836824697, + "grad_norm": 0.4524154406478645, + "learning_rate": 3.4204101075057026e-06, + "loss": 0.0785, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04045439511537552, + "step": 10540, + "valid_targets_mean": 3396.5, + "valid_targets_min": 1218 + }, + { + "epoch": 5.813120176405733, + "grad_norm": 0.41151516492513956, + "learning_rate": 3.4050510210352793e-06, + "loss": 0.0712, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04197981581091881, + "step": 10545, + "valid_targets_mean": 4247.9, + "valid_targets_min": 3261 + }, + { + "epoch": 5.815876515986769, + "grad_norm": 0.42202444969078207, + "learning_rate": 3.3897232872790675e-06, + "loss": 0.072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03484103083610535, + "step": 10550, + "valid_targets_mean": 2142.2, + "valid_targets_min": 719 + }, + { + "epoch": 5.818632855567806, + "grad_norm": 0.435871666056536, + "learning_rate": 3.374426935195643e-06, + "loss": 0.0686, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040170714259147644, + "step": 10555, + "valid_targets_mean": 3404.1, + "valid_targets_min": 2224 + }, + { + "epoch": 5.821389195148843, + "grad_norm": 0.32385619146670547, + "learning_rate": 3.3591619936842832e-06, + "loss": 0.0683, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.025465436279773712, + "step": 10560, + "valid_targets_mean": 3946.9, + "valid_targets_min": 3217 + }, + { + "epoch": 5.824145534729879, + "grad_norm": 0.46472559191320506, + "learning_rate": 3.34392849158494e-06, + "loss": 0.0695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03359426558017731, + "step": 10565, + "valid_targets_mean": 2405.0, + "valid_targets_min": 735 + }, + { + "epoch": 5.826901874310915, + "grad_norm": 0.4371572026305838, + "learning_rate": 3.328726457678153e-06, + "loss": 0.1034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030377404764294624, + "step": 10570, + "valid_targets_mean": 1502.2, + "valid_targets_min": 603 + }, + { + "epoch": 5.8296582138919515, + "grad_norm": 0.34353362648388497, + "learning_rate": 3.3135559206850075e-06, + "loss": 0.057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02294948510825634, + "step": 10575, + "valid_targets_mean": 1912.9, + "valid_targets_min": 582 + }, + { + "epoch": 5.832414553472988, + "grad_norm": 0.44266789512613053, + "learning_rate": 3.2984169092670927e-06, + "loss": 0.0715, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03873816132545471, + "step": 10580, + "valid_targets_mean": 2590.9, + "valid_targets_min": 579 + }, + { + "epoch": 5.835170893054024, + "grad_norm": 0.47547182366366675, + "learning_rate": 3.2833094520264265e-06, + "loss": 0.0644, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040246814489364624, + "step": 10585, + "valid_targets_mean": 3709.5, + "valid_targets_min": 3248 + }, + { + "epoch": 5.83792723263506, + "grad_norm": 0.43117213199161963, + "learning_rate": 3.2682335775054174e-06, + "loss": 0.1022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043621160089969635, + "step": 10590, + "valid_targets_mean": 3212.4, + "valid_targets_min": 994 + }, + { + "epoch": 5.840683572216097, + "grad_norm": 0.5712224197059349, + "learning_rate": 3.253189314186802e-06, + "loss": 0.0871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0644458681344986, + "step": 10595, + "valid_targets_mean": 2010.6, + "valid_targets_min": 692 + }, + { + "epoch": 5.843439911797134, + "grad_norm": 0.4548270706752876, + "learning_rate": 3.2381766904935997e-06, + "loss": 0.0884, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03507862612605095, + "step": 10600, + "valid_targets_mean": 2707.2, + "valid_targets_min": 632 + }, + { + "epoch": 5.84619625137817, + "grad_norm": 0.4491876289452511, + "learning_rate": 3.223195734789044e-06, + "loss": 0.0747, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033352792263031006, + "step": 10605, + "valid_targets_mean": 1847.8, + "valid_targets_min": 812 + }, + { + "epoch": 5.848952590959206, + "grad_norm": 0.5595339506335544, + "learning_rate": 3.208246475376542e-06, + "loss": 0.1193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14181698858737946, + "step": 10610, + "valid_targets_mean": 3245.8, + "valid_targets_min": 1177 + }, + { + "epoch": 5.8517089305402425, + "grad_norm": 0.4322615976517481, + "learning_rate": 3.193328940499629e-06, + "loss": 0.0806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038905978202819824, + "step": 10615, + "valid_targets_mean": 3384.0, + "valid_targets_min": 2331 + }, + { + "epoch": 5.854465270121279, + "grad_norm": 0.6371710484162666, + "learning_rate": 3.178443158341886e-06, + "loss": 0.1002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046945080161094666, + "step": 10620, + "valid_targets_mean": 1604.1, + "valid_targets_min": 684 + }, + { + "epoch": 5.857221609702315, + "grad_norm": 0.4860406424534109, + "learning_rate": 3.1635891570269115e-06, + "loss": 0.0682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026023222133517265, + "step": 10625, + "valid_targets_mean": 1534.4, + "valid_targets_min": 446 + }, + { + "epoch": 5.859977949283351, + "grad_norm": 0.3989391052205325, + "learning_rate": 3.148766964618262e-06, + "loss": 0.0771, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026056651026010513, + "step": 10630, + "valid_targets_mean": 2990.9, + "valid_targets_min": 655 + }, + { + "epoch": 5.862734288864388, + "grad_norm": 0.5901809663630798, + "learning_rate": 3.1339766091193937e-06, + "loss": 0.0817, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03634779527783394, + "step": 10635, + "valid_targets_mean": 1293.9, + "valid_targets_min": 574 + }, + { + "epoch": 5.865490628445425, + "grad_norm": 0.37210006066978457, + "learning_rate": 3.119218118473617e-06, + "loss": 0.066, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037052176892757416, + "step": 10640, + "valid_targets_mean": 4455.1, + "valid_targets_min": 2872 + }, + { + "epoch": 5.868246968026461, + "grad_norm": 0.4447077573210634, + "learning_rate": 3.104491520564039e-06, + "loss": 0.0623, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037745796144008636, + "step": 10645, + "valid_targets_mean": 3615.2, + "valid_targets_min": 2242 + }, + { + "epoch": 5.871003307607497, + "grad_norm": 0.4773826849576453, + "learning_rate": 3.0897968432135105e-06, + "loss": 0.0721, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03266705572605133, + "step": 10650, + "valid_targets_mean": 2017.4, + "valid_targets_min": 651 + }, + { + "epoch": 5.873759647188534, + "grad_norm": 0.4368175571318679, + "learning_rate": 3.0751341141845794e-06, + "loss": 0.0814, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03760548681020737, + "step": 10655, + "valid_targets_mean": 3131.8, + "valid_targets_min": 740 + }, + { + "epoch": 5.87651598676957, + "grad_norm": 0.46817477948874653, + "learning_rate": 3.060503361179423e-06, + "loss": 0.0848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04610780254006386, + "step": 10660, + "valid_targets_mean": 3651.6, + "valid_targets_min": 3006 + }, + { + "epoch": 5.879272326350606, + "grad_norm": 0.32251066237017467, + "learning_rate": 3.045904611839827e-06, + "loss": 0.075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02562575228512287, + "step": 10665, + "valid_targets_mean": 3182.6, + "valid_targets_min": 2353 + }, + { + "epoch": 5.882028665931642, + "grad_norm": 0.3744708112936466, + "learning_rate": 3.031337893747095e-06, + "loss": 0.0682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03274904191493988, + "step": 10670, + "valid_targets_mean": 1954.4, + "valid_targets_min": 665 + }, + { + "epoch": 5.884785005512679, + "grad_norm": 0.5449001038539208, + "learning_rate": 3.0168032344220165e-06, + "loss": 0.0845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04678473621606827, + "step": 10675, + "valid_targets_mean": 3512.6, + "valid_targets_min": 1776 + }, + { + "epoch": 5.887541345093716, + "grad_norm": 0.30783007072368207, + "learning_rate": 3.0023006613248217e-06, + "loss": 0.0593, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026593809947371483, + "step": 10680, + "valid_targets_mean": 4054.6, + "valid_targets_min": 3675 + }, + { + "epoch": 5.890297684674752, + "grad_norm": 0.3869931188894341, + "learning_rate": 2.9878302018551087e-06, + "loss": 0.076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030772458761930466, + "step": 10685, + "valid_targets_mean": 3238.4, + "valid_targets_min": 591 + }, + { + "epoch": 5.893054024255788, + "grad_norm": 0.4605642651115858, + "learning_rate": 2.9733918833518127e-06, + "loss": 0.0738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030945200473070145, + "step": 10690, + "valid_targets_mean": 2646.6, + "valid_targets_min": 1011 + }, + { + "epoch": 5.895810363836825, + "grad_norm": 0.6764040752373126, + "learning_rate": 2.9589857330931404e-06, + "loss": 0.0816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07497112452983856, + "step": 10695, + "valid_targets_mean": 1379.1, + "valid_targets_min": 903 + }, + { + "epoch": 5.8985667034178615, + "grad_norm": 0.5586774874598588, + "learning_rate": 2.9446117782965266e-06, + "loss": 0.0783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039083316922187805, + "step": 10700, + "valid_targets_mean": 2092.5, + "valid_targets_min": 693 + }, + { + "epoch": 5.901323042998897, + "grad_norm": 0.7043117433463113, + "learning_rate": 2.9302700461185707e-06, + "loss": 0.0748, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05158420652151108, + "step": 10705, + "valid_targets_mean": 1268.9, + "valid_targets_min": 732 + }, + { + "epoch": 5.904079382579933, + "grad_norm": 0.3905925919055544, + "learning_rate": 2.915960563655014e-06, + "loss": 0.0747, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02542797662317753, + "step": 10710, + "valid_targets_mean": 2524.1, + "valid_targets_min": 799 + }, + { + "epoch": 5.90683572216097, + "grad_norm": 0.48747317182592365, + "learning_rate": 2.9016833579406455e-06, + "loss": 0.094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03248729556798935, + "step": 10715, + "valid_targets_mean": 2966.2, + "valid_targets_min": 1012 + }, + { + "epoch": 5.909592061742007, + "grad_norm": 0.43850038214510684, + "learning_rate": 2.8874384559492897e-06, + "loss": 0.0752, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03640560805797577, + "step": 10720, + "valid_targets_mean": 3617.0, + "valid_targets_min": 1622 + }, + { + "epoch": 5.912348401323043, + "grad_norm": 0.4808803126567453, + "learning_rate": 2.8732258845937313e-06, + "loss": 0.0649, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036623515188694, + "step": 10725, + "valid_targets_mean": 3110.0, + "valid_targets_min": 1001 + }, + { + "epoch": 5.915104740904079, + "grad_norm": 0.5080290760899676, + "learning_rate": 2.8590456707256776e-06, + "loss": 0.0845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03609417751431465, + "step": 10730, + "valid_targets_mean": 2685.8, + "valid_targets_min": 897 + }, + { + "epoch": 5.917861080485116, + "grad_norm": 0.41242998779776086, + "learning_rate": 2.8448978411356987e-06, + "loss": 0.0795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02836194634437561, + "step": 10735, + "valid_targets_mean": 3736.1, + "valid_targets_min": 2915 + }, + { + "epoch": 5.9206174200661525, + "grad_norm": 0.5793019202099023, + "learning_rate": 2.830782422553189e-06, + "loss": 0.0625, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04963786527514458, + "step": 10740, + "valid_targets_mean": 2964.5, + "valid_targets_min": 561 + }, + { + "epoch": 5.9233737596471885, + "grad_norm": 0.4336134821873433, + "learning_rate": 2.8166994416463023e-06, + "loss": 0.0738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0413336344063282, + "step": 10745, + "valid_targets_mean": 3624.0, + "valid_targets_min": 2542 + }, + { + "epoch": 5.926130099228224, + "grad_norm": 0.33554586541828085, + "learning_rate": 2.802648925021907e-06, + "loss": 0.0645, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029299240559339523, + "step": 10750, + "valid_targets_mean": 4175.0, + "valid_targets_min": 2983 + }, + { + "epoch": 5.928886438809261, + "grad_norm": 0.4323167773218395, + "learning_rate": 2.788630899225542e-06, + "loss": 0.0599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026544589549303055, + "step": 10755, + "valid_targets_mean": 2350.6, + "valid_targets_min": 615 + }, + { + "epoch": 5.931642778390298, + "grad_norm": 0.33638575628980466, + "learning_rate": 2.7746453907413616e-06, + "loss": 0.0819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028164587914943695, + "step": 10760, + "valid_targets_mean": 3634.9, + "valid_targets_min": 2936 + }, + { + "epoch": 5.934399117971334, + "grad_norm": 0.35509146074754755, + "learning_rate": 2.760692425992084e-06, + "loss": 0.0729, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059672944247722626, + "step": 10765, + "valid_targets_mean": 3372.6, + "valid_targets_min": 1532 + }, + { + "epoch": 5.93715545755237, + "grad_norm": 0.351844325279397, + "learning_rate": 2.7467720313389402e-06, + "loss": 0.0879, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.022637424990534782, + "step": 10770, + "valid_targets_mean": 2305.8, + "valid_targets_min": 506 + }, + { + "epoch": 5.939911797133407, + "grad_norm": 0.33676799390919016, + "learning_rate": 2.73288423308163e-06, + "loss": 0.0809, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0225631482899189, + "step": 10775, + "valid_targets_mean": 2746.1, + "valid_targets_min": 633 + }, + { + "epoch": 5.942668136714444, + "grad_norm": 0.4439199814283108, + "learning_rate": 2.7190290574582647e-06, + "loss": 0.0939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039207179099321365, + "step": 10780, + "valid_targets_mean": 2730.4, + "valid_targets_min": 733 + }, + { + "epoch": 5.9454244762954795, + "grad_norm": 0.38883411985158495, + "learning_rate": 2.705206530645328e-06, + "loss": 0.0631, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032530527561903, + "step": 10785, + "valid_targets_mean": 3070.9, + "valid_targets_min": 489 + }, + { + "epoch": 5.948180815876516, + "grad_norm": 0.5298115403148507, + "learning_rate": 2.6914166787576167e-06, + "loss": 0.0787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030068688094615936, + "step": 10790, + "valid_targets_mean": 1371.4, + "valid_targets_min": 693 + }, + { + "epoch": 5.950937155457552, + "grad_norm": 0.40759207365693345, + "learning_rate": 2.6776595278481955e-06, + "loss": 0.0665, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027653295546770096, + "step": 10795, + "valid_targets_mean": 3525.9, + "valid_targets_min": 918 + }, + { + "epoch": 5.953693495038589, + "grad_norm": 0.6426167791812364, + "learning_rate": 2.6639351039083437e-06, + "loss": 0.097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03877793624997139, + "step": 10800, + "valid_targets_mean": 1479.6, + "valid_targets_min": 528 + }, + { + "epoch": 5.956449834619625, + "grad_norm": 0.5210138688689188, + "learning_rate": 2.650243432867521e-06, + "loss": 0.0918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05621990188956261, + "step": 10805, + "valid_targets_mean": 2083.9, + "valid_targets_min": 572 + }, + { + "epoch": 5.959206174200662, + "grad_norm": 0.4521832335046515, + "learning_rate": 2.636584540593299e-06, + "loss": 0.0719, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03792564198374748, + "step": 10810, + "valid_targets_mean": 3714.4, + "valid_targets_min": 2773 + }, + { + "epoch": 5.961962513781698, + "grad_norm": 0.3929132780614329, + "learning_rate": 2.62295845289132e-06, + "loss": 0.0703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03306751698255539, + "step": 10815, + "valid_targets_mean": 3400.9, + "valid_targets_min": 2395 + }, + { + "epoch": 5.964718853362735, + "grad_norm": 0.407902930840545, + "learning_rate": 2.6093651955052492e-06, + "loss": 0.0696, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027191732078790665, + "step": 10820, + "valid_targets_mean": 2494.9, + "valid_targets_min": 965 + }, + { + "epoch": 5.967475192943771, + "grad_norm": 0.6290227460414256, + "learning_rate": 2.5958047941167296e-06, + "loss": 0.0637, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02798750437796116, + "step": 10825, + "valid_targets_mean": 1522.5, + "valid_targets_min": 725 + }, + { + "epoch": 5.970231532524807, + "grad_norm": 0.9296617824979903, + "learning_rate": 2.582277274345324e-06, + "loss": 0.0826, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05650698393583298, + "step": 10830, + "valid_targets_mean": 1261.0, + "valid_targets_min": 633 + }, + { + "epoch": 5.972987872105843, + "grad_norm": 0.37082649663319855, + "learning_rate": 2.568782661748479e-06, + "loss": 0.0629, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030363786965608597, + "step": 10835, + "valid_targets_mean": 3052.9, + "valid_targets_min": 889 + }, + { + "epoch": 5.97574421168688, + "grad_norm": 0.3825294583797634, + "learning_rate": 2.5553209818214632e-06, + "loss": 0.0667, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030417509377002716, + "step": 10840, + "valid_targets_mean": 3525.2, + "valid_targets_min": 2282 + }, + { + "epoch": 5.978500551267916, + "grad_norm": 0.4593179671302999, + "learning_rate": 2.541892259997332e-06, + "loss": 0.0664, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03162161633372307, + "step": 10845, + "valid_targets_mean": 3317.0, + "valid_targets_min": 981 + }, + { + "epoch": 5.981256890848953, + "grad_norm": 0.3838053120083722, + "learning_rate": 2.5284965216468617e-06, + "loss": 0.0764, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03357156738638878, + "step": 10850, + "valid_targets_mean": 3465.1, + "valid_targets_min": 1071 + }, + { + "epoch": 5.984013230429989, + "grad_norm": 0.37351271588893603, + "learning_rate": 2.5151337920785323e-06, + "loss": 0.0624, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030290085822343826, + "step": 10855, + "valid_targets_mean": 4006.9, + "valid_targets_min": 757 + }, + { + "epoch": 5.986769570011026, + "grad_norm": 0.36973672066951335, + "learning_rate": 2.501804096538447e-06, + "loss": 0.0676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0293755941092968, + "step": 10860, + "valid_targets_mean": 1548.0, + "valid_targets_min": 504 + }, + { + "epoch": 5.989525909592062, + "grad_norm": 0.4800377622497491, + "learning_rate": 2.4885074602102987e-06, + "loss": 0.0838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07441917061805725, + "step": 10865, + "valid_targets_mean": 2691.9, + "valid_targets_min": 775 + }, + { + "epoch": 5.9922822491730985, + "grad_norm": 0.3996715414143398, + "learning_rate": 2.4752439082153235e-06, + "loss": 0.1384, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03516942262649536, + "step": 10870, + "valid_targets_mean": 3880.9, + "valid_targets_min": 2338 + }, + { + "epoch": 5.995038588754134, + "grad_norm": 0.45355554616555877, + "learning_rate": 2.462013465612254e-06, + "loss": 0.0683, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03833349421620369, + "step": 10875, + "valid_targets_mean": 3656.0, + "valid_targets_min": 2463 + }, + { + "epoch": 5.997794928335171, + "grad_norm": 0.3829625410006069, + "learning_rate": 2.4488161573972667e-06, + "loss": 0.0668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031292110681533813, + "step": 10880, + "valid_targets_mean": 3262.0, + "valid_targets_min": 2237 + }, + { + "epoch": 6.000551267916207, + "grad_norm": 0.8802162946209984, + "learning_rate": 2.435652008503939e-06, + "loss": 0.121, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09622689336538315, + "step": 10885, + "valid_targets_mean": 8186.1, + "valid_targets_min": 6131 + }, + { + "epoch": 6.003307607497244, + "grad_norm": 0.8511095389490271, + "learning_rate": 2.422521043803199e-06, + "loss": 0.1894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0983312577009201, + "step": 10890, + "valid_targets_mean": 8564.9, + "valid_targets_min": 5974 + }, + { + "epoch": 6.00606394707828, + "grad_norm": 0.7528307150013224, + "learning_rate": 2.4094232881032787e-06, + "loss": 0.1723, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07573948800563812, + "step": 10895, + "valid_targets_mean": 3567.1, + "valid_targets_min": 287 + }, + { + "epoch": 6.008820286659317, + "grad_norm": 0.6140655918477461, + "learning_rate": 2.3963587661496713e-06, + "loss": 0.174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08844809234142303, + "step": 10900, + "valid_targets_mean": 6991.4, + "valid_targets_min": 4953 + }, + { + "epoch": 6.011576626240353, + "grad_norm": 0.5186704917595246, + "learning_rate": 2.383327502625088e-06, + "loss": 0.1641, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08503441512584686, + "step": 10905, + "valid_targets_mean": 6929.9, + "valid_targets_min": 1449 + }, + { + "epoch": 6.0143329658213895, + "grad_norm": 0.45431074681047334, + "learning_rate": 2.3703295221493926e-06, + "loss": 0.1673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07884803414344788, + "step": 10910, + "valid_targets_mean": 7335.5, + "valid_targets_min": 4961 + }, + { + "epoch": 6.0170893054024255, + "grad_norm": 0.6099601464271528, + "learning_rate": 2.3573648492795754e-06, + "loss": 0.1808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10537317395210266, + "step": 10915, + "valid_targets_mean": 7028.9, + "valid_targets_min": 5952 + }, + { + "epoch": 6.019845644983462, + "grad_norm": 0.45359479295224625, + "learning_rate": 2.3444335085096957e-06, + "loss": 0.1752, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09445784986019135, + "step": 10920, + "valid_targets_mean": 7634.2, + "valid_targets_min": 5143 + }, + { + "epoch": 6.022601984564498, + "grad_norm": 0.4545238090972308, + "learning_rate": 2.33153552427084e-06, + "loss": 0.1643, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08367142826318741, + "step": 10925, + "valid_targets_mean": 7079.8, + "valid_targets_min": 5316 + }, + { + "epoch": 6.025358324145535, + "grad_norm": 0.48384796454008555, + "learning_rate": 2.318670920931074e-06, + "loss": 0.1525, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0696815699338913, + "step": 10930, + "valid_targets_mean": 4218.1, + "valid_targets_min": 151 + }, + { + "epoch": 6.028114663726571, + "grad_norm": 0.47878082583634385, + "learning_rate": 2.3058397227954e-06, + "loss": 0.1569, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08404131978750229, + "step": 10935, + "valid_targets_mean": 7171.6, + "valid_targets_min": 4244 + }, + { + "epoch": 6.030871003307608, + "grad_norm": 0.3930177638332718, + "learning_rate": 2.2930419541057015e-06, + "loss": 0.1653, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07873982191085815, + "step": 10940, + "valid_targets_mean": 7214.5, + "valid_targets_min": 5799 + }, + { + "epoch": 6.033627342888644, + "grad_norm": 0.4007690383936676, + "learning_rate": 2.2802776390407087e-06, + "loss": 0.1617, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08218531310558319, + "step": 10945, + "valid_targets_mean": 6923.1, + "valid_targets_min": 3844 + }, + { + "epoch": 6.036383682469681, + "grad_norm": 0.3928059243958965, + "learning_rate": 2.2675468017159518e-06, + "loss": 0.1616, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08643710613250732, + "step": 10950, + "valid_targets_mean": 6961.0, + "valid_targets_min": 5664 + }, + { + "epoch": 6.0391400220507165, + "grad_norm": 0.39669451206706696, + "learning_rate": 2.2548494661837063e-06, + "loss": 0.1544, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08575719594955444, + "step": 10955, + "valid_targets_mean": 6933.6, + "valid_targets_min": 4915 + }, + { + "epoch": 6.041896361631753, + "grad_norm": 0.39662153027625346, + "learning_rate": 2.2421856564329515e-06, + "loss": 0.16, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07464984059333801, + "step": 10960, + "valid_targets_mean": 6407.2, + "valid_targets_min": 4966 + }, + { + "epoch": 6.044652701212789, + "grad_norm": 0.3929320658499919, + "learning_rate": 2.2295553963893313e-06, + "loss": 0.1616, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0793992131948471, + "step": 10965, + "valid_targets_mean": 7489.0, + "valid_targets_min": 5451 + }, + { + "epoch": 6.047409040793826, + "grad_norm": 0.3700913877814535, + "learning_rate": 2.2169587099151002e-06, + "loss": 0.157, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07545287162065506, + "step": 10970, + "valid_targets_mean": 7392.1, + "valid_targets_min": 5195 + }, + { + "epoch": 6.050165380374862, + "grad_norm": 0.43129561426113106, + "learning_rate": 2.204395620809088e-06, + "loss": 0.1676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10293616354465485, + "step": 10975, + "valid_targets_mean": 7050.0, + "valid_targets_min": 3733 + }, + { + "epoch": 6.052921719955899, + "grad_norm": 0.5071941691046834, + "learning_rate": 2.191866152806643e-06, + "loss": 0.1854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09210155159235, + "step": 10980, + "valid_targets_mean": 6038.4, + "valid_targets_min": 5143 + }, + { + "epoch": 6.055678059536935, + "grad_norm": 0.39194477379696757, + "learning_rate": 2.1793703295795998e-06, + "loss": 0.1291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07496733963489532, + "step": 10985, + "valid_targets_mean": 7356.9, + "valid_targets_min": 5765 + }, + { + "epoch": 6.058434399117972, + "grad_norm": 0.4318048175364649, + "learning_rate": 2.1669081747362196e-06, + "loss": 0.1532, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07312336564064026, + "step": 10990, + "valid_targets_mean": 7668.2, + "valid_targets_min": 4640 + }, + { + "epoch": 6.061190738699008, + "grad_norm": 0.4149528731585041, + "learning_rate": 2.1544797118211604e-06, + "loss": 0.1581, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08484707027673721, + "step": 10995, + "valid_targets_mean": 7758.0, + "valid_targets_min": 5483 + }, + { + "epoch": 6.063947078280044, + "grad_norm": 0.41776973048587823, + "learning_rate": 2.1420849643154294e-06, + "loss": 0.1646, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08095023036003113, + "step": 11000, + "valid_targets_mean": 7311.8, + "valid_targets_min": 6424 + }, + { + "epoch": 6.06670341786108, + "grad_norm": 0.4268776459874466, + "learning_rate": 2.129723955636329e-06, + "loss": 0.1597, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07035191357135773, + "step": 11005, + "valid_targets_mean": 6788.4, + "valid_targets_min": 4371 + }, + { + "epoch": 6.069459757442117, + "grad_norm": 0.3726134409337157, + "learning_rate": 2.1173967091374227e-06, + "loss": 0.1593, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06995213031768799, + "step": 11010, + "valid_targets_mean": 7830.8, + "valid_targets_min": 5371 + }, + { + "epoch": 6.072216097023153, + "grad_norm": 0.35189671466657413, + "learning_rate": 2.105103248108482e-06, + "loss": 0.1397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06630485504865646, + "step": 11015, + "valid_targets_mean": 8286.8, + "valid_targets_min": 4818 + }, + { + "epoch": 6.07497243660419, + "grad_norm": 0.3522325805506894, + "learning_rate": 2.0928435957754557e-06, + "loss": 0.1413, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0631987601518631, + "step": 11020, + "valid_targets_mean": 7339.1, + "valid_targets_min": 5995 + }, + { + "epoch": 6.077728776185226, + "grad_norm": 0.443594213462284, + "learning_rate": 2.080617775300413e-06, + "loss": 0.1469, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0791100561618805, + "step": 11025, + "valid_targets_mean": 6405.4, + "valid_targets_min": 4206 + }, + { + "epoch": 6.080485115766263, + "grad_norm": 1.125044085873397, + "learning_rate": 2.0684258097815047e-06, + "loss": 0.1426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0382051058113575, + "step": 11030, + "valid_targets_mean": 246.4, + "valid_targets_min": 136 + }, + { + "epoch": 6.083241455347299, + "grad_norm": 0.40699652176806905, + "learning_rate": 2.0562677222529204e-06, + "loss": 0.1636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0778571367263794, + "step": 11035, + "valid_targets_mean": 6778.4, + "valid_targets_min": 5133 + }, + { + "epoch": 6.0859977949283355, + "grad_norm": 0.48125059161722583, + "learning_rate": 2.0441435356848436e-06, + "loss": 0.1704, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0880143940448761, + "step": 11040, + "valid_targets_mean": 6963.0, + "valid_targets_min": 5072 + }, + { + "epoch": 6.088754134509371, + "grad_norm": 0.4491228283912413, + "learning_rate": 2.0320532729834187e-06, + "loss": 0.1668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09027764201164246, + "step": 11045, + "valid_targets_mean": 6662.5, + "valid_targets_min": 5036 + }, + { + "epoch": 6.091510474090408, + "grad_norm": 0.45685641968402607, + "learning_rate": 2.019996956990682e-06, + "loss": 0.165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09951740503311157, + "step": 11050, + "valid_targets_mean": 7389.0, + "valid_targets_min": 5546 + }, + { + "epoch": 6.094266813671444, + "grad_norm": 0.7172622326394562, + "learning_rate": 2.0079746104845445e-06, + "loss": 0.1589, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08385463058948517, + "step": 11055, + "valid_targets_mean": 7507.0, + "valid_targets_min": 5051 + }, + { + "epoch": 6.097023153252481, + "grad_norm": 0.365469255758816, + "learning_rate": 1.9959862561787347e-06, + "loss": 0.1647, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.074983149766922, + "step": 11060, + "valid_targets_mean": 7192.6, + "valid_targets_min": 5809 + }, + { + "epoch": 6.099779492833517, + "grad_norm": 0.38834490207464745, + "learning_rate": 1.984031916722762e-06, + "loss": 0.1596, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07787619531154633, + "step": 11065, + "valid_targets_mean": 7255.8, + "valid_targets_min": 5119 + }, + { + "epoch": 6.102535832414554, + "grad_norm": 0.4170106191843786, + "learning_rate": 1.972111614701873e-06, + "loss": 0.1685, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07847284525632858, + "step": 11070, + "valid_targets_mean": 6710.9, + "valid_targets_min": 4029 + }, + { + "epoch": 6.10529217199559, + "grad_norm": 0.41976871771617336, + "learning_rate": 1.9602253726370056e-06, + "loss": 0.1629, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09080667793750763, + "step": 11075, + "valid_targets_mean": 6299.6, + "valid_targets_min": 5464 + }, + { + "epoch": 6.1080485115766265, + "grad_norm": 0.41581867759370217, + "learning_rate": 1.9483732129847466e-06, + "loss": 0.1634, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0784960389137268, + "step": 11080, + "valid_targets_mean": 6239.8, + "valid_targets_min": 4663 + }, + { + "epoch": 6.1108048511576625, + "grad_norm": 0.4839678782426013, + "learning_rate": 1.9365551581372964e-06, + "loss": 0.1615, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08362806588411331, + "step": 11085, + "valid_targets_mean": 5759.1, + "valid_targets_min": 4989 + }, + { + "epoch": 6.113561190738699, + "grad_norm": 0.4791406587691499, + "learning_rate": 1.924771230422413e-06, + "loss": 0.1537, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05924566835165024, + "step": 11090, + "valid_targets_mean": 3368.0, + "valid_targets_min": 2264 + }, + { + "epoch": 6.116317530319735, + "grad_norm": 0.3991121484280943, + "learning_rate": 1.913021452103392e-06, + "loss": 0.1315, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07860555499792099, + "step": 11095, + "valid_targets_mean": 6208.5, + "valid_targets_min": 5061 + }, + { + "epoch": 6.119073869900772, + "grad_norm": 0.4406699925615219, + "learning_rate": 1.901305845378998e-06, + "loss": 0.1579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08362892270088196, + "step": 11100, + "valid_targets_mean": 6299.4, + "valid_targets_min": 4566 + }, + { + "epoch": 6.121830209481808, + "grad_norm": 0.3990293253587393, + "learning_rate": 1.8896244323834412e-06, + "loss": 0.1578, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08541477471590042, + "step": 11105, + "valid_targets_mean": 7560.9, + "valid_targets_min": 4931 + }, + { + "epoch": 6.124586549062845, + "grad_norm": 0.4267325555239708, + "learning_rate": 1.877977235186328e-06, + "loss": 0.1602, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06893756985664368, + "step": 11110, + "valid_targets_mean": 5546.4, + "valid_targets_min": 4426 + }, + { + "epoch": 6.127342888643881, + "grad_norm": 0.4420573001834005, + "learning_rate": 1.8663642757926227e-06, + "loss": 0.1576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07653450965881348, + "step": 11115, + "valid_targets_mean": 5400.6, + "valid_targets_min": 4401 + }, + { + "epoch": 6.130099228224918, + "grad_norm": 0.44871912343287074, + "learning_rate": 1.8547855761426014e-06, + "loss": 0.1613, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0828271210193634, + "step": 11120, + "valid_targets_mean": 6330.9, + "valid_targets_min": 5178 + }, + { + "epoch": 6.1328555678059535, + "grad_norm": 0.42350946604276296, + "learning_rate": 1.843241158111817e-06, + "loss": 0.1614, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07641628384590149, + "step": 11125, + "valid_targets_mean": 5793.9, + "valid_targets_min": 5097 + }, + { + "epoch": 6.13561190738699, + "grad_norm": 0.807233614050785, + "learning_rate": 1.8317310435110535e-06, + "loss": 0.1222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07271047681570053, + "step": 11130, + "valid_targets_mean": 1627.8, + "valid_targets_min": 136 + }, + { + "epoch": 6.138368246968026, + "grad_norm": 0.4124063597926261, + "learning_rate": 1.820255254086285e-06, + "loss": 0.1545, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0797586590051651, + "step": 11135, + "valid_targets_mean": 6442.6, + "valid_targets_min": 4719 + }, + { + "epoch": 6.141124586549063, + "grad_norm": 0.43132827385954997, + "learning_rate": 1.8088138115186304e-06, + "loss": 0.1561, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08573848009109497, + "step": 11140, + "valid_targets_mean": 7150.1, + "valid_targets_min": 5326 + }, + { + "epoch": 6.143880926130099, + "grad_norm": 0.40334199508534185, + "learning_rate": 1.7974067374243298e-06, + "loss": 0.161, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07998141646385193, + "step": 11145, + "valid_targets_mean": 7101.2, + "valid_targets_min": 4056 + }, + { + "epoch": 6.146637265711136, + "grad_norm": 0.41493274883508985, + "learning_rate": 1.786034053354684e-06, + "loss": 0.1534, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07590934634208679, + "step": 11150, + "valid_targets_mean": 6992.9, + "valid_targets_min": 4903 + }, + { + "epoch": 6.149393605292172, + "grad_norm": 0.44747635115251494, + "learning_rate": 1.7746957807960141e-06, + "loss": 0.1503, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08159563690423965, + "step": 11155, + "valid_targets_mean": 5869.9, + "valid_targets_min": 4747 + }, + { + "epoch": 6.152149944873209, + "grad_norm": 0.4058747345945952, + "learning_rate": 1.763391941169641e-06, + "loss": 0.1579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08126654475927353, + "step": 11160, + "valid_targets_mean": 6258.1, + "valid_targets_min": 4588 + }, + { + "epoch": 6.154906284454245, + "grad_norm": 0.8679267113020909, + "learning_rate": 1.7521225558318212e-06, + "loss": 0.1463, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053155090659856796, + "step": 11165, + "valid_targets_mean": 924.5, + "valid_targets_min": 533 + }, + { + "epoch": 6.157662624035281, + "grad_norm": 0.8584263844016579, + "learning_rate": 1.7408876460737234e-06, + "loss": 0.1334, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056692831218242645, + "step": 11170, + "valid_targets_mean": 1198.2, + "valid_targets_min": 688 + }, + { + "epoch": 6.160418963616317, + "grad_norm": 0.7257602456363798, + "learning_rate": 1.7296872331213776e-06, + "loss": 0.1262, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05788172408938408, + "step": 11175, + "valid_targets_mean": 1214.4, + "valid_targets_min": 655 + }, + { + "epoch": 6.163175303197354, + "grad_norm": 0.7549807896862138, + "learning_rate": 1.7185213381356414e-06, + "loss": 0.1205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057677045464515686, + "step": 11180, + "valid_targets_mean": 1183.0, + "valid_targets_min": 579 + }, + { + "epoch": 6.16593164277839, + "grad_norm": 0.7355064785464688, + "learning_rate": 1.7073899822121532e-06, + "loss": 0.1225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057982079684734344, + "step": 11185, + "valid_targets_mean": 1302.1, + "valid_targets_min": 494 + }, + { + "epoch": 6.168687982359427, + "grad_norm": 0.7134327550104151, + "learning_rate": 1.6962931863813103e-06, + "loss": 0.1168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06083889305591583, + "step": 11190, + "valid_targets_mean": 1632.4, + "valid_targets_min": 723 + }, + { + "epoch": 6.171444321940463, + "grad_norm": 0.7491545456260879, + "learning_rate": 1.6852309716082027e-06, + "loss": 0.1228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04521341994404793, + "step": 11195, + "valid_targets_mean": 1045.1, + "valid_targets_min": 631 + }, + { + "epoch": 6.1742006615215, + "grad_norm": 0.8368656508019763, + "learning_rate": 1.6742033587925921e-06, + "loss": 0.1097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06716330349445343, + "step": 11200, + "valid_targets_mean": 1272.4, + "valid_targets_min": 711 + }, + { + "epoch": 6.176957001102536, + "grad_norm": 0.8029994682078684, + "learning_rate": 1.6632103687688617e-06, + "loss": 0.1331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052628643810749054, + "step": 11205, + "valid_targets_mean": 1190.1, + "valid_targets_min": 649 + }, + { + "epoch": 6.1797133406835725, + "grad_norm": 0.8074490573103726, + "learning_rate": 1.652252022305989e-06, + "loss": 0.124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07343283295631409, + "step": 11210, + "valid_targets_mean": 1823.4, + "valid_targets_min": 788 + }, + { + "epoch": 6.182469680264608, + "grad_norm": 0.7997674426984371, + "learning_rate": 1.6413283401074975e-06, + "loss": 0.1114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05735182389616966, + "step": 11215, + "valid_targets_mean": 1319.8, + "valid_targets_min": 552 + }, + { + "epoch": 6.185226019845645, + "grad_norm": 0.7950694189675335, + "learning_rate": 1.6304393428114185e-06, + "loss": 0.1145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06400343030691147, + "step": 11220, + "valid_targets_mean": 1626.4, + "valid_targets_min": 1005 + }, + { + "epoch": 6.187982359426681, + "grad_norm": 0.7502145496874684, + "learning_rate": 1.6195850509902534e-06, + "loss": 0.1162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04769829660654068, + "step": 11225, + "valid_targets_mean": 1193.9, + "valid_targets_min": 495 + }, + { + "epoch": 6.190738699007718, + "grad_norm": 0.8093075360248628, + "learning_rate": 1.6087654851509339e-06, + "loss": 0.1017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04602108895778656, + "step": 11230, + "valid_targets_mean": 1010.6, + "valid_targets_min": 538 + }, + { + "epoch": 6.193495038588754, + "grad_norm": 0.7680826334028743, + "learning_rate": 1.597980665734784e-06, + "loss": 0.1026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05260644480586052, + "step": 11235, + "valid_targets_mean": 1310.4, + "valid_targets_min": 639 + }, + { + "epoch": 6.196251378169791, + "grad_norm": 0.7913246854897845, + "learning_rate": 1.5872306131174896e-06, + "loss": 0.106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037261031568050385, + "step": 11240, + "valid_targets_mean": 914.5, + "valid_targets_min": 540 + }, + { + "epoch": 6.199007717750827, + "grad_norm": 0.9007909093975396, + "learning_rate": 1.5765153476090444e-06, + "loss": 0.1143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05050533264875412, + "step": 11245, + "valid_targets_mean": 1156.4, + "valid_targets_min": 629 + }, + { + "epoch": 6.2017640573318635, + "grad_norm": 0.952292051413419, + "learning_rate": 1.5658348894537167e-06, + "loss": 0.1179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07495544850826263, + "step": 11250, + "valid_targets_mean": 1455.2, + "valid_targets_min": 506 + }, + { + "epoch": 6.2045203969128995, + "grad_norm": 0.8296839978400403, + "learning_rate": 1.5551892588300188e-06, + "loss": 0.1043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050474874675273895, + "step": 11255, + "valid_targets_mean": 1249.9, + "valid_targets_min": 553 + }, + { + "epoch": 6.207276736493936, + "grad_norm": 0.8123411690147391, + "learning_rate": 1.5445784758506599e-06, + "loss": 0.1159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04946425184607506, + "step": 11260, + "valid_targets_mean": 1130.6, + "valid_targets_min": 488 + }, + { + "epoch": 6.210033076074972, + "grad_norm": 0.893088708904895, + "learning_rate": 1.5340025605625152e-06, + "loss": 0.115, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05851132422685623, + "step": 11265, + "valid_targets_mean": 1354.8, + "valid_targets_min": 620 + }, + { + "epoch": 6.212789415656009, + "grad_norm": 0.8699721086540261, + "learning_rate": 1.523461532946584e-06, + "loss": 0.1136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05165359377861023, + "step": 11270, + "valid_targets_mean": 1189.6, + "valid_targets_min": 690 + }, + { + "epoch": 6.215545755237045, + "grad_norm": 0.8435336795913779, + "learning_rate": 1.5129554129179514e-06, + "loss": 0.1084, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027229661121964455, + "step": 11275, + "valid_targets_mean": 658.8, + "valid_targets_min": 460 + }, + { + "epoch": 6.218302094818082, + "grad_norm": 0.9074824545906341, + "learning_rate": 1.5024842203257506e-06, + "loss": 0.1056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05499300733208656, + "step": 11280, + "valid_targets_mean": 1305.6, + "valid_targets_min": 499 + }, + { + "epoch": 6.221058434399118, + "grad_norm": 0.7753737084932779, + "learning_rate": 1.4920479749531302e-06, + "loss": 0.1076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06278610229492188, + "step": 11285, + "valid_targets_mean": 1593.8, + "valid_targets_min": 561 + }, + { + "epoch": 6.223814773980155, + "grad_norm": 0.8098094879961589, + "learning_rate": 1.4816466965172094e-06, + "loss": 0.1142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05979905650019646, + "step": 11290, + "valid_targets_mean": 1541.1, + "valid_targets_min": 1057 + }, + { + "epoch": 6.2265711135611905, + "grad_norm": 0.8817911746722572, + "learning_rate": 1.4712804046690466e-06, + "loss": 0.1145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0683225691318512, + "step": 11295, + "valid_targets_mean": 1733.4, + "valid_targets_min": 1103 + }, + { + "epoch": 6.229327453142227, + "grad_norm": 0.9265884348042271, + "learning_rate": 1.460949118993602e-06, + "loss": 0.1152, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055870186537504196, + "step": 11300, + "valid_targets_mean": 1164.4, + "valid_targets_min": 635 + }, + { + "epoch": 6.232083792723263, + "grad_norm": 0.8413792183671707, + "learning_rate": 1.4506528590096957e-06, + "loss": 0.1061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0731881856918335, + "step": 11305, + "valid_targets_mean": 1747.9, + "valid_targets_min": 786 + }, + { + "epoch": 6.2348401323043, + "grad_norm": 0.8724640645309992, + "learning_rate": 1.4403916441699762e-06, + "loss": 0.1098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05792447179555893, + "step": 11310, + "valid_targets_mean": 1471.6, + "valid_targets_min": 697 + }, + { + "epoch": 6.237596471885336, + "grad_norm": 0.8628383294854215, + "learning_rate": 1.43016549386088e-06, + "loss": 0.1059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05639030411839485, + "step": 11315, + "valid_targets_mean": 1413.9, + "valid_targets_min": 612 + }, + { + "epoch": 6.240352811466373, + "grad_norm": 0.8596742457176973, + "learning_rate": 1.419974427402595e-06, + "loss": 0.1106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05761212110519409, + "step": 11320, + "valid_targets_mean": 1360.5, + "valid_targets_min": 1015 + }, + { + "epoch": 6.243109151047409, + "grad_norm": 0.9024806747480316, + "learning_rate": 1.4098184640490331e-06, + "loss": 0.1061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03708692640066147, + "step": 11325, + "valid_targets_mean": 810.5, + "valid_targets_min": 521 + }, + { + "epoch": 6.245865490628446, + "grad_norm": 0.839500842515692, + "learning_rate": 1.3996976229877768e-06, + "loss": 0.1021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059936653822660446, + "step": 11330, + "valid_targets_mean": 1617.4, + "valid_targets_min": 815 + }, + { + "epoch": 6.248621830209482, + "grad_norm": 1.036278191971053, + "learning_rate": 1.3896119233400595e-06, + "loss": 0.0981, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05496181547641754, + "step": 11335, + "valid_targets_mean": 1233.5, + "valid_targets_min": 605 + }, + { + "epoch": 6.251378169790518, + "grad_norm": 0.8888861864973994, + "learning_rate": 1.3795613841607148e-06, + "loss": 0.1062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03669361025094986, + "step": 11340, + "valid_targets_mean": 799.8, + "valid_targets_min": 478 + }, + { + "epoch": 6.254134509371554, + "grad_norm": 0.7821918954507052, + "learning_rate": 1.369546024438153e-06, + "loss": 0.1047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04804329574108124, + "step": 11345, + "valid_targets_mean": 1218.1, + "valid_targets_min": 522 + }, + { + "epoch": 6.256890848952591, + "grad_norm": 0.7959614771060475, + "learning_rate": 1.3595658630943186e-06, + "loss": 0.0986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04443267732858658, + "step": 11350, + "valid_targets_mean": 1148.9, + "valid_targets_min": 539 + }, + { + "epoch": 6.259647188533627, + "grad_norm": 0.8797350475867082, + "learning_rate": 1.34962091898466e-06, + "loss": 0.1064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06606210768222809, + "step": 11355, + "valid_targets_mean": 1526.0, + "valid_targets_min": 539 + }, + { + "epoch": 6.262403528114664, + "grad_norm": 0.9573545230993225, + "learning_rate": 1.3397112108980826e-06, + "loss": 0.1105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06634832918643951, + "step": 11360, + "valid_targets_mean": 1506.6, + "valid_targets_min": 1024 + }, + { + "epoch": 6.2651598676957, + "grad_norm": 0.8999152119184359, + "learning_rate": 1.329836757556926e-06, + "loss": 0.1014, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06468009948730469, + "step": 11365, + "valid_targets_mean": 1525.4, + "valid_targets_min": 579 + }, + { + "epoch": 6.267916207276737, + "grad_norm": 0.9670301287547244, + "learning_rate": 1.3199975776169183e-06, + "loss": 0.1089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06408125907182693, + "step": 11370, + "valid_targets_mean": 1283.8, + "valid_targets_min": 693 + }, + { + "epoch": 6.270672546857773, + "grad_norm": 0.9134562575696672, + "learning_rate": 1.3101936896671562e-06, + "loss": 0.1127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05560162663459778, + "step": 11375, + "valid_targets_mean": 1233.0, + "valid_targets_min": 944 + }, + { + "epoch": 6.2734288864388095, + "grad_norm": 0.8639500309308136, + "learning_rate": 1.3004251122300481e-06, + "loss": 0.1146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054976433515548706, + "step": 11380, + "valid_targets_mean": 1430.8, + "valid_targets_min": 471 + }, + { + "epoch": 6.276185226019845, + "grad_norm": 0.8502502135745069, + "learning_rate": 1.2906918637612975e-06, + "loss": 0.101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05893056094646454, + "step": 11385, + "valid_targets_mean": 1406.9, + "valid_targets_min": 626 + }, + { + "epoch": 6.278941565600882, + "grad_norm": 0.9733979297881021, + "learning_rate": 1.2809939626498547e-06, + "loss": 0.1068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059218280017375946, + "step": 11390, + "valid_targets_mean": 1598.9, + "valid_targets_min": 793 + }, + { + "epoch": 6.281697905181918, + "grad_norm": 0.8890284018837907, + "learning_rate": 1.2713314272178967e-06, + "loss": 0.1053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050279274582862854, + "step": 11395, + "valid_targets_mean": 1102.5, + "valid_targets_min": 546 + }, + { + "epoch": 6.284454244762955, + "grad_norm": 0.8535745306984652, + "learning_rate": 1.2617042757207787e-06, + "loss": 0.0962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041645653545856476, + "step": 11400, + "valid_targets_mean": 960.6, + "valid_targets_min": 463 + }, + { + "epoch": 6.287210584343991, + "grad_norm": 0.8509327710530309, + "learning_rate": 1.2521125263470046e-06, + "loss": 0.1037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05614414066076279, + "step": 11405, + "valid_targets_mean": 1476.2, + "valid_targets_min": 502 + }, + { + "epoch": 6.289966923925028, + "grad_norm": 0.8426840775258699, + "learning_rate": 1.2425561972181987e-06, + "loss": 0.1061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058310218155384064, + "step": 11410, + "valid_targets_mean": 1639.6, + "valid_targets_min": 791 + }, + { + "epoch": 6.292723263506064, + "grad_norm": 0.8208415506835091, + "learning_rate": 1.2330353063890588e-06, + "loss": 0.1027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03948543220758438, + "step": 11415, + "valid_targets_mean": 1105.2, + "valid_targets_min": 650 + }, + { + "epoch": 6.2954796030871005, + "grad_norm": 0.8779900587411062, + "learning_rate": 1.223549871847336e-06, + "loss": 0.1086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0413164347410202, + "step": 11420, + "valid_targets_mean": 1122.5, + "valid_targets_min": 783 + }, + { + "epoch": 6.2982359426681365, + "grad_norm": 0.8797772822498073, + "learning_rate": 1.2140999115137953e-06, + "loss": 0.1006, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042214758694171906, + "step": 11425, + "valid_targets_mean": 1240.1, + "valid_targets_min": 594 + }, + { + "epoch": 6.300992282249173, + "grad_norm": 0.8748846213887059, + "learning_rate": 1.2046854432421706e-06, + "loss": 0.1043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04717801511287689, + "step": 11430, + "valid_targets_mean": 1214.1, + "valid_targets_min": 599 + }, + { + "epoch": 6.303748621830209, + "grad_norm": 0.9039652539387826, + "learning_rate": 1.1953064848191497e-06, + "loss": 0.1063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06343744695186615, + "step": 11435, + "valid_targets_mean": 1340.0, + "valid_targets_min": 708 + }, + { + "epoch": 6.306504961411246, + "grad_norm": 0.8522649011579951, + "learning_rate": 1.1859630539643297e-06, + "loss": 0.1075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04055285453796387, + "step": 11440, + "valid_targets_mean": 949.5, + "valid_targets_min": 626 + }, + { + "epoch": 6.309261300992282, + "grad_norm": 0.8942730959078462, + "learning_rate": 1.176655168330183e-06, + "loss": 0.1083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055376820266246796, + "step": 11445, + "valid_targets_mean": 1736.5, + "valid_targets_min": 940 + }, + { + "epoch": 6.312017640573319, + "grad_norm": 0.9267190620181753, + "learning_rate": 1.167382845502032e-06, + "loss": 0.0985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044667430222034454, + "step": 11450, + "valid_targets_mean": 1206.4, + "valid_targets_min": 641 + }, + { + "epoch": 6.314773980154355, + "grad_norm": 0.9337690087378936, + "learning_rate": 1.1581461029980078e-06, + "loss": 0.108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0595521554350853, + "step": 11455, + "valid_targets_mean": 1205.4, + "valid_targets_min": 492 + }, + { + "epoch": 6.317530319735392, + "grad_norm": 0.8598682070808191, + "learning_rate": 1.148944958269016e-06, + "loss": 0.1128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036048293113708496, + "step": 11460, + "valid_targets_mean": 1165.6, + "valid_targets_min": 592 + }, + { + "epoch": 6.3202866593164275, + "grad_norm": 0.885427526167507, + "learning_rate": 1.1397794286987107e-06, + "loss": 0.1004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03821457922458649, + "step": 11465, + "valid_targets_mean": 831.5, + "valid_targets_min": 533 + }, + { + "epoch": 6.323042998897464, + "grad_norm": 0.8368497413532319, + "learning_rate": 1.1306495316034648e-06, + "loss": 0.1028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0642569363117218, + "step": 11470, + "valid_targets_mean": 1535.8, + "valid_targets_min": 1072 + }, + { + "epoch": 6.3257993384785, + "grad_norm": 0.8382724002524439, + "learning_rate": 1.1215552842323208e-06, + "loss": 0.1073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047637976706027985, + "step": 11475, + "valid_targets_mean": 1162.1, + "valid_targets_min": 576 + }, + { + "epoch": 6.328555678059537, + "grad_norm": 0.8697228554992293, + "learning_rate": 1.1124967037669764e-06, + "loss": 0.1095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05076310783624649, + "step": 11480, + "valid_targets_mean": 1351.5, + "valid_targets_min": 505 + }, + { + "epoch": 6.331312017640573, + "grad_norm": 0.9893618305029528, + "learning_rate": 1.1034738073217356e-06, + "loss": 0.109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057002242654561996, + "step": 11485, + "valid_targets_mean": 1197.6, + "valid_targets_min": 747 + }, + { + "epoch": 6.33406835722161, + "grad_norm": 0.8273461967125143, + "learning_rate": 1.0944866119434948e-06, + "loss": 0.1018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05162757635116577, + "step": 11490, + "valid_targets_mean": 1411.5, + "valid_targets_min": 1107 + }, + { + "epoch": 6.336824696802646, + "grad_norm": 0.8555669185451902, + "learning_rate": 1.0855351346116927e-06, + "loss": 0.1099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04792550951242447, + "step": 11495, + "valid_targets_mean": 1218.2, + "valid_targets_min": 579 + }, + { + "epoch": 6.339581036383683, + "grad_norm": 0.9657922143622871, + "learning_rate": 1.0766193922382895e-06, + "loss": 0.1021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05176892131567001, + "step": 11500, + "valid_targets_mean": 1226.9, + "valid_targets_min": 535 + }, + { + "epoch": 6.342337375964719, + "grad_norm": 0.9402216379751431, + "learning_rate": 1.0677394016677333e-06, + "loss": 0.1104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06545155495405197, + "step": 11505, + "valid_targets_mean": 1649.1, + "valid_targets_min": 610 + }, + { + "epoch": 6.345093715545755, + "grad_norm": 0.9299043640933735, + "learning_rate": 1.058895179676922e-06, + "loss": 0.1157, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04715527966618538, + "step": 11510, + "valid_targets_mean": 1128.8, + "valid_targets_min": 720 + }, + { + "epoch": 6.347850055126791, + "grad_norm": 0.9954104122187878, + "learning_rate": 1.050086742975176e-06, + "loss": 0.1087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060182634741067886, + "step": 11515, + "valid_targets_mean": 1602.5, + "valid_targets_min": 1013 + }, + { + "epoch": 6.350606394707828, + "grad_norm": 0.8428058084437697, + "learning_rate": 1.0413141082042167e-06, + "loss": 0.1031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034302107989788055, + "step": 11520, + "valid_targets_mean": 712.9, + "valid_targets_min": 465 + }, + { + "epoch": 6.353362734288864, + "grad_norm": 0.8963169415452714, + "learning_rate": 1.0325772919381127e-06, + "loss": 0.1038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045319072902202606, + "step": 11525, + "valid_targets_mean": 1248.9, + "valid_targets_min": 465 + }, + { + "epoch": 6.356119073869901, + "grad_norm": 0.8376791825308839, + "learning_rate": 1.0238763106832694e-06, + "loss": 0.1057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04126331955194473, + "step": 11530, + "valid_targets_mean": 1212.0, + "valid_targets_min": 527 + }, + { + "epoch": 6.358875413450937, + "grad_norm": 0.8664141845460533, + "learning_rate": 1.015211180878386e-06, + "loss": 0.103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051978714764118195, + "step": 11535, + "valid_targets_mean": 1509.4, + "valid_targets_min": 561 + }, + { + "epoch": 6.361631753031974, + "grad_norm": 0.8236292128493155, + "learning_rate": 1.0065819188944247e-06, + "loss": 0.0984, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04504565894603729, + "step": 11540, + "valid_targets_mean": 1309.4, + "valid_targets_min": 466 + }, + { + "epoch": 6.36438809261301, + "grad_norm": 0.8441645119141639, + "learning_rate": 9.97988541034589e-07, + "loss": 0.1061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05205211043357849, + "step": 11545, + "valid_targets_mean": 1201.4, + "valid_targets_min": 574 + }, + { + "epoch": 6.3671444321940465, + "grad_norm": 0.8489797102227278, + "learning_rate": 9.894310635342852e-07, + "loss": 0.1145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04767122492194176, + "step": 11550, + "valid_targets_mean": 1237.0, + "valid_targets_min": 929 + }, + { + "epoch": 6.369900771775082, + "grad_norm": 0.8812423048040239, + "learning_rate": 9.809095025610894e-07, + "loss": 0.1081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05183082073926926, + "step": 11555, + "valid_targets_mean": 1202.0, + "valid_targets_min": 732 + }, + { + "epoch": 6.372657111356119, + "grad_norm": 0.7584087787093254, + "learning_rate": 9.724238742147207e-07, + "loss": 0.1021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042624980211257935, + "step": 11560, + "valid_targets_mean": 1180.6, + "valid_targets_min": 581 + }, + { + "epoch": 6.375413450937155, + "grad_norm": 0.7521470422746602, + "learning_rate": 9.63974194527022e-07, + "loss": 0.1027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05408954247832298, + "step": 11565, + "valid_targets_mean": 1458.8, + "valid_targets_min": 1090 + }, + { + "epoch": 6.378169790518192, + "grad_norm": 0.9302139791010817, + "learning_rate": 9.555604794619056e-07, + "loss": 0.094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03922674059867859, + "step": 11570, + "valid_targets_mean": 907.9, + "valid_targets_min": 468 + }, + { + "epoch": 6.380926130099228, + "grad_norm": 0.8538200977107541, + "learning_rate": 9.471827449153425e-07, + "loss": 0.1034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04976292699575424, + "step": 11575, + "valid_targets_mean": 1355.2, + "valid_targets_min": 492 + }, + { + "epoch": 6.383682469680265, + "grad_norm": 0.9174697055360336, + "learning_rate": 9.38841006715323e-07, + "loss": 0.1023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053636834025382996, + "step": 11580, + "valid_targets_mean": 1405.4, + "valid_targets_min": 704 + }, + { + "epoch": 6.386438809261301, + "grad_norm": 0.8733978817632125, + "learning_rate": 9.305352806218338e-07, + "loss": 0.1062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046556949615478516, + "step": 11585, + "valid_targets_mean": 1271.2, + "valid_targets_min": 551 + }, + { + "epoch": 6.3891951488423375, + "grad_norm": 0.8111896336516622, + "learning_rate": 9.222655823268201e-07, + "loss": 0.1067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051444657146930695, + "step": 11590, + "valid_targets_mean": 1294.6, + "valid_targets_min": 582 + }, + { + "epoch": 6.3919514884233735, + "grad_norm": 0.8435057271216753, + "learning_rate": 9.140319274541621e-07, + "loss": 0.0993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04381068795919418, + "step": 11595, + "valid_targets_mean": 1146.5, + "valid_targets_min": 637 + }, + { + "epoch": 6.39470782800441, + "grad_norm": 0.9027976849928646, + "learning_rate": 9.058343315596407e-07, + "loss": 0.1044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052246250212192535, + "step": 11600, + "valid_targets_mean": 1311.0, + "valid_targets_min": 911 + }, + { + "epoch": 6.397464167585446, + "grad_norm": 1.0320510650450618, + "learning_rate": 8.976728101309162e-07, + "loss": 0.1941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10400396585464478, + "step": 11605, + "valid_targets_mean": 4293.2, + "valid_targets_min": 464 + }, + { + "epoch": 6.400220507166483, + "grad_norm": 1.0367861027686596, + "learning_rate": 8.895473785874875e-07, + "loss": 0.215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12209682166576385, + "step": 11610, + "valid_targets_mean": 4942.9, + "valid_targets_min": 551 + }, + { + "epoch": 6.402976846747519, + "grad_norm": 0.8958156926253719, + "learning_rate": 8.814580522806748e-07, + "loss": 0.2138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08480393886566162, + "step": 11615, + "valid_targets_mean": 4082.9, + "valid_targets_min": 1827 + }, + { + "epoch": 6.405733186328556, + "grad_norm": 0.9524638184588685, + "learning_rate": 8.734048464935841e-07, + "loss": 0.2068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08631639927625656, + "step": 11620, + "valid_targets_mean": 3299.1, + "valid_targets_min": 1441 + }, + { + "epoch": 6.408489525909592, + "grad_norm": 0.8451722709048037, + "learning_rate": 8.653877764410734e-07, + "loss": 0.2061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1276102364063263, + "step": 11625, + "valid_targets_mean": 4508.8, + "valid_targets_min": 1547 + }, + { + "epoch": 6.411245865490629, + "grad_norm": 0.7518115081391414, + "learning_rate": 8.574068572697359e-07, + "loss": 0.2104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10504645854234695, + "step": 11630, + "valid_targets_mean": 5931.4, + "valid_targets_min": 1511 + }, + { + "epoch": 6.4140022050716645, + "grad_norm": 0.8298533445769729, + "learning_rate": 8.494621040578632e-07, + "loss": 0.2, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11721840500831604, + "step": 11635, + "valid_targets_mean": 7490.9, + "valid_targets_min": 2040 + }, + { + "epoch": 6.416758544652701, + "grad_norm": 0.7001030955687733, + "learning_rate": 8.415535318154178e-07, + "loss": 0.2024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06535481661558151, + "step": 11640, + "valid_targets_mean": 2932.9, + "valid_targets_min": 290 + }, + { + "epoch": 6.419514884233737, + "grad_norm": 0.7160611952965399, + "learning_rate": 8.336811554840074e-07, + "loss": 0.2083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11884547770023346, + "step": 11645, + "valid_targets_mean": 5701.0, + "valid_targets_min": 1290 + }, + { + "epoch": 6.422271223814774, + "grad_norm": 0.6888025962427632, + "learning_rate": 8.258449899368526e-07, + "loss": 0.2068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09789913147687912, + "step": 11650, + "valid_targets_mean": 4178.8, + "valid_targets_min": 1054 + }, + { + "epoch": 6.42502756339581, + "grad_norm": 0.7294560941515934, + "learning_rate": 8.180450499787662e-07, + "loss": 0.2026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0995173528790474, + "step": 11655, + "valid_targets_mean": 3651.5, + "valid_targets_min": 968 + }, + { + "epoch": 6.427783902976847, + "grad_norm": 0.6347279719285082, + "learning_rate": 8.102813503461115e-07, + "loss": 0.202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10698485374450684, + "step": 11660, + "valid_targets_mean": 5212.2, + "valid_targets_min": 1273 + }, + { + "epoch": 6.430540242557883, + "grad_norm": 0.6654513415396901, + "learning_rate": 8.025539057067954e-07, + "loss": 0.2013, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10321217775344849, + "step": 11665, + "valid_targets_mean": 3800.2, + "valid_targets_min": 956 + }, + { + "epoch": 6.43329658213892, + "grad_norm": 0.6575971440146096, + "learning_rate": 7.948627306602174e-07, + "loss": 0.1942, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1001708060503006, + "step": 11670, + "valid_targets_mean": 4258.0, + "valid_targets_min": 1441 + }, + { + "epoch": 6.436052921719956, + "grad_norm": 0.6643225920781176, + "learning_rate": 7.872078397372607e-07, + "loss": 0.1949, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08097740262746811, + "step": 11675, + "valid_targets_mean": 3077.5, + "valid_targets_min": 924 + }, + { + "epoch": 6.438809261300992, + "grad_norm": 0.6861022114517445, + "learning_rate": 7.795892474002497e-07, + "loss": 0.1902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09778115898370743, + "step": 11680, + "valid_targets_mean": 3072.4, + "valid_targets_min": 1407 + }, + { + "epoch": 6.441565600882028, + "grad_norm": 0.7947807412183985, + "learning_rate": 7.720069680429399e-07, + "loss": 0.1937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0932692289352417, + "step": 11685, + "valid_targets_mean": 2772.9, + "valid_targets_min": 382 + }, + { + "epoch": 6.444321940463065, + "grad_norm": 0.6834728425395336, + "learning_rate": 7.644610159904719e-07, + "loss": 0.1961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10433506965637207, + "step": 11690, + "valid_targets_mean": 3333.5, + "valid_targets_min": 1021 + }, + { + "epoch": 6.447078280044101, + "grad_norm": 0.7579215872613594, + "learning_rate": 7.569514054993576e-07, + "loss": 0.1911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07272063195705414, + "step": 11695, + "valid_targets_mean": 1783.8, + "valid_targets_min": 541 + }, + { + "epoch": 6.449834619625138, + "grad_norm": 0.6468088282275044, + "learning_rate": 7.494781507574478e-07, + "loss": 0.1964, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09279416501522064, + "step": 11700, + "valid_targets_mean": 3691.0, + "valid_targets_min": 1207 + }, + { + "epoch": 6.452590959206174, + "grad_norm": 0.7304325301498936, + "learning_rate": 7.42041265883906e-07, + "loss": 0.1922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08798564970493317, + "step": 11705, + "valid_targets_mean": 3065.6, + "valid_targets_min": 767 + }, + { + "epoch": 6.455347298787211, + "grad_norm": 0.7237002471925252, + "learning_rate": 7.346407649291887e-07, + "loss": 0.1859, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12483751028776169, + "step": 11710, + "valid_targets_mean": 3770.9, + "valid_targets_min": 1544 + }, + { + "epoch": 6.458103638368247, + "grad_norm": 0.6112262468652957, + "learning_rate": 7.272766618750049e-07, + "loss": 0.171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07426966726779938, + "step": 11715, + "valid_targets_mean": 3893.1, + "valid_targets_min": 1633 + }, + { + "epoch": 6.4608599779492835, + "grad_norm": 0.6890717631620322, + "learning_rate": 7.19948970634301e-07, + "loss": 0.185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08869637548923492, + "step": 11720, + "valid_targets_mean": 3006.1, + "valid_targets_min": 1381 + }, + { + "epoch": 6.463616317530319, + "grad_norm": 0.7087201709304558, + "learning_rate": 7.126577050512274e-07, + "loss": 0.1936, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09885524958372116, + "step": 11725, + "valid_targets_mean": 3179.0, + "valid_targets_min": 418 + }, + { + "epoch": 6.466372657111356, + "grad_norm": 0.6336620434486898, + "learning_rate": 7.05402878901118e-07, + "loss": 0.1923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09644827991724014, + "step": 11730, + "valid_targets_mean": 3881.4, + "valid_targets_min": 1596 + }, + { + "epoch": 6.469128996692392, + "grad_norm": 0.7144455870692364, + "learning_rate": 6.981845058904646e-07, + "loss": 0.1768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07082241028547287, + "step": 11735, + "valid_targets_mean": 2007.9, + "valid_targets_min": 994 + }, + { + "epoch": 6.471885336273429, + "grad_norm": 0.6408913912743396, + "learning_rate": 6.91002599656887e-07, + "loss": 0.1812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09001576900482178, + "step": 11740, + "valid_targets_mean": 3667.6, + "valid_targets_min": 1738 + }, + { + "epoch": 6.474641675854465, + "grad_norm": 0.6890666516836838, + "learning_rate": 6.838571737691024e-07, + "loss": 0.1902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09241239726543427, + "step": 11745, + "valid_targets_mean": 3595.1, + "valid_targets_min": 612 + }, + { + "epoch": 6.477398015435502, + "grad_norm": 0.6753158446790869, + "learning_rate": 6.767482417269145e-07, + "loss": 0.182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08299222588539124, + "step": 11750, + "valid_targets_mean": 2771.0, + "valid_targets_min": 1190 + }, + { + "epoch": 6.480154355016538, + "grad_norm": 0.6126445980658085, + "learning_rate": 6.69675816961175e-07, + "loss": 0.1721, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0754721462726593, + "step": 11755, + "valid_targets_mean": 4476.0, + "valid_targets_min": 1648 + }, + { + "epoch": 6.4829106945975745, + "grad_norm": 0.673660825440815, + "learning_rate": 6.626399128337647e-07, + "loss": 0.1877, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08212882280349731, + "step": 11760, + "valid_targets_mean": 3159.1, + "valid_targets_min": 951 + }, + { + "epoch": 6.4856670341786105, + "grad_norm": 0.6285040876432009, + "learning_rate": 6.556405426375634e-07, + "loss": 0.1811, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09563671797513962, + "step": 11765, + "valid_targets_mean": 3956.6, + "valid_targets_min": 1863 + }, + { + "epoch": 6.488423373759647, + "grad_norm": 0.7053813702701041, + "learning_rate": 6.486777195964334e-07, + "loss": 0.1756, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10518098622560501, + "step": 11770, + "valid_targets_mean": 3484.0, + "valid_targets_min": 2153 + }, + { + "epoch": 6.491179713340683, + "grad_norm": 0.7011677969611494, + "learning_rate": 6.417514568651783e-07, + "loss": 0.1807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0801987498998642, + "step": 11775, + "valid_targets_mean": 2673.0, + "valid_targets_min": 1094 + }, + { + "epoch": 6.49393605292172, + "grad_norm": 0.6624985835182912, + "learning_rate": 6.348617675295399e-07, + "loss": 0.1817, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08723919093608856, + "step": 11780, + "valid_targets_mean": 3583.8, + "valid_targets_min": 885 + }, + { + "epoch": 6.496692392502756, + "grad_norm": 0.5829534546956378, + "learning_rate": 6.280086646061567e-07, + "loss": 0.1849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0980951339006424, + "step": 11785, + "valid_targets_mean": 5273.2, + "valid_targets_min": 1686 + }, + { + "epoch": 6.499448732083793, + "grad_norm": 0.6524290290233319, + "learning_rate": 6.211921610425431e-07, + "loss": 0.1747, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09701443463563919, + "step": 11790, + "valid_targets_mean": 3796.8, + "valid_targets_min": 1682 + }, + { + "epoch": 6.502205071664829, + "grad_norm": 0.794111754680152, + "learning_rate": 6.144122697170707e-07, + "loss": 0.1861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0986676812171936, + "step": 11795, + "valid_targets_mean": 3432.5, + "valid_targets_min": 1640 + }, + { + "epoch": 6.504961411245866, + "grad_norm": 0.7102964128354835, + "learning_rate": 6.076690034389355e-07, + "loss": 0.1883, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12555021047592163, + "step": 11800, + "valid_targets_mean": 4120.2, + "valid_targets_min": 700 + }, + { + "epoch": 6.5077177508269015, + "grad_norm": 0.7108847476774399, + "learning_rate": 6.009623749481442e-07, + "loss": 0.1733, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09560355544090271, + "step": 11805, + "valid_targets_mean": 3153.4, + "valid_targets_min": 577 + }, + { + "epoch": 6.510474090407938, + "grad_norm": 0.6710043140947429, + "learning_rate": 5.94292396915479e-07, + "loss": 0.1846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11344876885414124, + "step": 11810, + "valid_targets_mean": 4260.0, + "valid_targets_min": 2219 + }, + { + "epoch": 6.513230429988974, + "grad_norm": 0.668497393651786, + "learning_rate": 5.876590819424777e-07, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10497400164604187, + "step": 11815, + "valid_targets_mean": 4229.1, + "valid_targets_min": 1825 + }, + { + "epoch": 6.515986769570011, + "grad_norm": 0.6549051936322039, + "learning_rate": 5.810624425614175e-07, + "loss": 0.173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09451582282781601, + "step": 11820, + "valid_targets_mean": 3777.9, + "valid_targets_min": 2010 + }, + { + "epoch": 6.518743109151047, + "grad_norm": 0.6305377032969298, + "learning_rate": 5.745024912352759e-07, + "loss": 0.1716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056851692497730255, + "step": 11825, + "valid_targets_mean": 2502.8, + "valid_targets_min": 734 + }, + { + "epoch": 6.521499448732084, + "grad_norm": 0.8049612788222872, + "learning_rate": 5.679792403577234e-07, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11676248908042908, + "step": 11830, + "valid_targets_mean": 3351.5, + "valid_targets_min": 1697 + }, + { + "epoch": 6.52425578831312, + "grad_norm": 0.7343429680989126, + "learning_rate": 5.614927022530881e-07, + "loss": 0.1829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07836558669805527, + "step": 11835, + "valid_targets_mean": 2532.9, + "valid_targets_min": 1377 + }, + { + "epoch": 6.527012127894157, + "grad_norm": 0.739816907584939, + "learning_rate": 5.550428891763382e-07, + "loss": 0.1779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10288634896278381, + "step": 11840, + "valid_targets_mean": 3187.2, + "valid_targets_min": 2106 + }, + { + "epoch": 6.529768467475193, + "grad_norm": 0.665689957532118, + "learning_rate": 5.486298133130574e-07, + "loss": 0.1876, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09983654320240021, + "step": 11845, + "valid_targets_mean": 3775.5, + "valid_targets_min": 1896 + }, + { + "epoch": 6.532524807056229, + "grad_norm": 0.6858287613679602, + "learning_rate": 5.422534867794227e-07, + "loss": 0.1828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12415570020675659, + "step": 11850, + "valid_targets_mean": 4860.4, + "valid_targets_min": 1760 + }, + { + "epoch": 6.535281146637265, + "grad_norm": 0.6846989339586901, + "learning_rate": 5.359139216221843e-07, + "loss": 0.1795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11500412970781326, + "step": 11855, + "valid_targets_mean": 3965.0, + "valid_targets_min": 1967 + }, + { + "epoch": 6.538037486218302, + "grad_norm": 0.6891123057927732, + "learning_rate": 5.296111298186324e-07, + "loss": 0.1893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12977728247642517, + "step": 11860, + "valid_targets_mean": 4136.6, + "valid_targets_min": 1647 + }, + { + "epoch": 6.540793825799338, + "grad_norm": 0.6763509181536684, + "learning_rate": 5.23345123276584e-07, + "loss": 0.1715, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07689844071865082, + "step": 11865, + "valid_targets_mean": 3405.9, + "valid_targets_min": 1902 + }, + { + "epoch": 6.543550165380375, + "grad_norm": 0.6785851298608492, + "learning_rate": 5.171159138343607e-07, + "loss": 0.1739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10189792513847351, + "step": 11870, + "valid_targets_mean": 3465.0, + "valid_targets_min": 2261 + }, + { + "epoch": 6.546306504961411, + "grad_norm": 0.5860171120004964, + "learning_rate": 5.109235132607593e-07, + "loss": 0.1725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07792720943689346, + "step": 11875, + "valid_targets_mean": 4239.8, + "valid_targets_min": 902 + }, + { + "epoch": 6.549062844542448, + "grad_norm": 0.7652463304679279, + "learning_rate": 5.047679332550392e-07, + "loss": 0.183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10939890146255493, + "step": 11880, + "valid_targets_mean": 3418.5, + "valid_targets_min": 742 + }, + { + "epoch": 6.551819184123484, + "grad_norm": 0.6606862368826837, + "learning_rate": 4.986491854468933e-07, + "loss": 0.1838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09033980965614319, + "step": 11885, + "valid_targets_mean": 3846.0, + "valid_targets_min": 1355 + }, + { + "epoch": 6.5545755237045205, + "grad_norm": 0.7378912076359447, + "learning_rate": 4.925672813964211e-07, + "loss": 0.1864, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09526233375072479, + "step": 11890, + "valid_targets_mean": 3270.1, + "valid_targets_min": 1242 + }, + { + "epoch": 6.557331863285556, + "grad_norm": 0.7771548879424613, + "learning_rate": 4.865222325941244e-07, + "loss": 0.1814, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07633332908153534, + "step": 11895, + "valid_targets_mean": 3767.4, + "valid_targets_min": 2073 + }, + { + "epoch": 6.560088202866593, + "grad_norm": 0.6611231253598411, + "learning_rate": 4.805140504608629e-07, + "loss": 0.1836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08433166146278381, + "step": 11900, + "valid_targets_mean": 3248.2, + "valid_targets_min": 906 + }, + { + "epoch": 6.56284454244763, + "grad_norm": 0.6211008096651486, + "learning_rate": 4.7454274634785914e-07, + "loss": 0.1863, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0770096480846405, + "step": 11905, + "valid_targets_mean": 3214.2, + "valid_targets_min": 616 + }, + { + "epoch": 6.565600882028666, + "grad_norm": 0.7277036359206339, + "learning_rate": 4.686083315366463e-07, + "loss": 0.18, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07547272741794586, + "step": 11910, + "valid_targets_mean": 2117.8, + "valid_targets_min": 819 + }, + { + "epoch": 6.568357221609702, + "grad_norm": 0.6933556910599639, + "learning_rate": 4.627108172390737e-07, + "loss": 0.1756, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09703750908374786, + "step": 11915, + "valid_targets_mean": 3294.6, + "valid_targets_min": 1143 + }, + { + "epoch": 6.571113561190739, + "grad_norm": 0.6519953390882741, + "learning_rate": 4.5685021459727305e-07, + "loss": 0.1659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0876879170536995, + "step": 11920, + "valid_targets_mean": 4368.0, + "valid_targets_min": 1191 + }, + { + "epoch": 6.573869900771776, + "grad_norm": 0.6493625576561036, + "learning_rate": 4.510265346836318e-07, + "loss": 0.1675, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09066452085971832, + "step": 11925, + "valid_targets_mean": 3497.1, + "valid_targets_min": 289 + }, + { + "epoch": 6.5766262403528115, + "grad_norm": 0.6702275192839398, + "learning_rate": 4.4523978850079087e-07, + "loss": 0.1659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0659290999174118, + "step": 11930, + "valid_targets_mean": 3379.6, + "valid_targets_min": 989 + }, + { + "epoch": 6.5793825799338475, + "grad_norm": 0.7381833960238823, + "learning_rate": 4.3948998698160496e-07, + "loss": 0.1689, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07123301178216934, + "step": 11935, + "valid_targets_mean": 1870.4, + "valid_targets_min": 344 + }, + { + "epoch": 6.582138919514884, + "grad_norm": 0.7215116794218666, + "learning_rate": 4.3377714098913115e-07, + "loss": 0.1646, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09057755768299103, + "step": 11940, + "valid_targets_mean": 3110.2, + "valid_targets_min": 2053 + }, + { + "epoch": 6.584895259095921, + "grad_norm": 0.7220840889533369, + "learning_rate": 4.281012613166069e-07, + "loss": 0.1803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13213279843330383, + "step": 11945, + "valid_targets_mean": 4253.2, + "valid_targets_min": 2066 + }, + { + "epoch": 6.587651598676957, + "grad_norm": 0.6643996077538709, + "learning_rate": 4.2246235868742993e-07, + "loss": 0.1881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07527261972427368, + "step": 11950, + "valid_targets_mean": 3006.6, + "valid_targets_min": 1502 + }, + { + "epoch": 6.590407938257993, + "grad_norm": 0.8025928494481409, + "learning_rate": 4.1686044375513824e-07, + "loss": 0.1701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09493574500083923, + "step": 11955, + "valid_targets_mean": 2874.1, + "valid_targets_min": 800 + }, + { + "epoch": 6.59316427783903, + "grad_norm": 0.6603801563580183, + "learning_rate": 4.1129552710338806e-07, + "loss": 0.1786, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0802958533167839, + "step": 11960, + "valid_targets_mean": 3433.6, + "valid_targets_min": 1415 + }, + { + "epoch": 6.595920617420067, + "grad_norm": 0.6366445141980711, + "learning_rate": 4.0576761924593364e-07, + "loss": 0.1711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09964495152235031, + "step": 11965, + "valid_targets_mean": 4436.1, + "valid_targets_min": 1292 + }, + { + "epoch": 6.598676957001103, + "grad_norm": 0.5854263195773312, + "learning_rate": 4.00276730626612e-07, + "loss": 0.1712, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07669363915920258, + "step": 11970, + "valid_targets_mean": 5184.6, + "valid_targets_min": 1521 + }, + { + "epoch": 6.6014332965821385, + "grad_norm": 0.5842100429837755, + "learning_rate": 3.948228716193181e-07, + "loss": 0.1763, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08000318706035614, + "step": 11975, + "valid_targets_mean": 4194.6, + "valid_targets_min": 1789 + }, + { + "epoch": 6.604189636163175, + "grad_norm": 0.7855397764768485, + "learning_rate": 3.8940605252798525e-07, + "loss": 0.1693, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07417282462120056, + "step": 11980, + "valid_targets_mean": 2431.6, + "valid_targets_min": 1470 + }, + { + "epoch": 6.606945975744212, + "grad_norm": 0.6133515407517024, + "learning_rate": 3.840262835865716e-07, + "loss": 0.1791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06169552728533745, + "step": 11985, + "valid_targets_mean": 2841.1, + "valid_targets_min": 1282 + }, + { + "epoch": 6.609702315325248, + "grad_norm": 0.6742099317070862, + "learning_rate": 3.7868357495903563e-07, + "loss": 0.169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09125593304634094, + "step": 11990, + "valid_targets_mean": 3547.6, + "valid_targets_min": 1422 + }, + { + "epoch": 6.612458654906284, + "grad_norm": 0.701458012834621, + "learning_rate": 3.7337793673931423e-07, + "loss": 0.1766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08068627864122391, + "step": 11995, + "valid_targets_mean": 2877.2, + "valid_targets_min": 1210 + }, + { + "epoch": 6.615214994487321, + "grad_norm": 0.7695848302389805, + "learning_rate": 3.681093789513135e-07, + "loss": 0.1835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0887434333562851, + "step": 12000, + "valid_targets_mean": 3072.8, + "valid_targets_min": 1139 + }, + { + "epoch": 6.617971334068358, + "grad_norm": 0.8244715162664032, + "learning_rate": 3.6287791154887785e-07, + "loss": 0.1768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09874193370342255, + "step": 12005, + "valid_targets_mean": 2247.0, + "valid_targets_min": 619 + }, + { + "epoch": 6.620727673649394, + "grad_norm": 0.6774782173895799, + "learning_rate": 3.5768354441578333e-07, + "loss": 0.1897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06477059423923492, + "step": 12010, + "valid_targets_mean": 2174.6, + "valid_targets_min": 447 + }, + { + "epoch": 6.62348401323043, + "grad_norm": 0.6201300225097001, + "learning_rate": 3.525262873657065e-07, + "loss": 0.1716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0917324423789978, + "step": 12015, + "valid_targets_mean": 4156.4, + "valid_targets_min": 2452 + }, + { + "epoch": 6.626240352811466, + "grad_norm": 0.6829642598728204, + "learning_rate": 3.474061501422155e-07, + "loss": 0.1909, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07371805608272552, + "step": 12020, + "valid_targets_mean": 3091.2, + "valid_targets_min": 837 + }, + { + "epoch": 6.628996692392503, + "grad_norm": 0.7691955657768629, + "learning_rate": 3.423231424187501e-07, + "loss": 0.1851, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11921127885580063, + "step": 12025, + "valid_targets_mean": 3410.5, + "valid_targets_min": 794 + }, + { + "epoch": 6.631753031973539, + "grad_norm": 0.7296700120493347, + "learning_rate": 3.3727727379859523e-07, + "loss": 0.1738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09127651900053024, + "step": 12030, + "valid_targets_mean": 3586.6, + "valid_targets_min": 808 + }, + { + "epoch": 6.634509371554575, + "grad_norm": 0.6845525899671606, + "learning_rate": 3.32268553814874e-07, + "loss": 0.1762, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09687060117721558, + "step": 12035, + "valid_targets_mean": 3692.8, + "valid_targets_min": 1948 + }, + { + "epoch": 6.637265711135612, + "grad_norm": 0.6771746853633223, + "learning_rate": 3.2729699193052357e-07, + "loss": 0.1768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07548873126506805, + "step": 12040, + "valid_targets_mean": 3656.9, + "valid_targets_min": 1909 + }, + { + "epoch": 6.640022050716649, + "grad_norm": 0.7244509069592122, + "learning_rate": 3.223625975382838e-07, + "loss": 0.1788, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08296230435371399, + "step": 12045, + "valid_targets_mean": 2981.6, + "valid_targets_min": 1511 + }, + { + "epoch": 6.642778390297685, + "grad_norm": 0.7817593895650017, + "learning_rate": 3.174653799606642e-07, + "loss": 0.1824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08907763659954071, + "step": 12050, + "valid_targets_mean": 2484.5, + "valid_targets_min": 450 + }, + { + "epoch": 6.645534729878721, + "grad_norm": 0.6234383539378722, + "learning_rate": 3.126053484499436e-07, + "loss": 0.1634, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07648061215877533, + "step": 12055, + "valid_targets_mean": 3432.0, + "valid_targets_min": 2554 + }, + { + "epoch": 6.6482910694597575, + "grad_norm": 0.6527591176254739, + "learning_rate": 3.0778251218814394e-07, + "loss": 0.1725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08831062912940979, + "step": 12060, + "valid_targets_mean": 3475.6, + "valid_targets_min": 1202 + }, + { + "epoch": 6.651047409040794, + "grad_norm": 0.6841593227106207, + "learning_rate": 3.0299688028701204e-07, + "loss": 0.1709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08239395171403885, + "step": 12065, + "valid_targets_mean": 3600.8, + "valid_targets_min": 2159 + }, + { + "epoch": 6.65380374862183, + "grad_norm": 0.836515335223071, + "learning_rate": 2.9824846178801105e-07, + "loss": 0.1696, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0859474390745163, + "step": 12070, + "valid_targets_mean": 2471.4, + "valid_targets_min": 808 + }, + { + "epoch": 6.656560088202866, + "grad_norm": 0.5990868010432479, + "learning_rate": 2.93537265662287e-07, + "loss": 0.1738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07420231401920319, + "step": 12075, + "valid_targets_mean": 3015.4, + "valid_targets_min": 990 + }, + { + "epoch": 6.659316427783903, + "grad_norm": 0.7099057248873211, + "learning_rate": 2.888633008106734e-07, + "loss": 0.1778, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0866604745388031, + "step": 12080, + "valid_targets_mean": 3459.4, + "valid_targets_min": 318 + }, + { + "epoch": 6.66207276736494, + "grad_norm": 3.6546277644653844, + "learning_rate": 2.842265760636531e-07, + "loss": 0.1729, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0958932489156723, + "step": 12085, + "valid_targets_mean": 4340.6, + "valid_targets_min": 2582 + }, + { + "epoch": 6.664829106945976, + "grad_norm": 0.7347543490767082, + "learning_rate": 2.796271001813566e-07, + "loss": 0.1728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09001035988330841, + "step": 12090, + "valid_targets_mean": 2988.0, + "valid_targets_min": 1117 + }, + { + "epoch": 6.667585446527012, + "grad_norm": 0.6061585992557804, + "learning_rate": 2.75064881853544e-07, + "loss": 0.1818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07674999535083771, + "step": 12095, + "valid_targets_mean": 3749.8, + "valid_targets_min": 1890 + }, + { + "epoch": 6.6703417861080485, + "grad_norm": 0.6808117098023337, + "learning_rate": 2.705399296995759e-07, + "loss": 0.1721, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07426311075687408, + "step": 12100, + "valid_targets_mean": 2966.2, + "valid_targets_min": 1793 + }, + { + "epoch": 6.673098125689085, + "grad_norm": 0.6399353171589107, + "learning_rate": 2.660522522684117e-07, + "loss": 0.1625, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07246260344982147, + "step": 12105, + "valid_targets_mean": 3508.5, + "valid_targets_min": 773 + }, + { + "epoch": 6.675854465270121, + "grad_norm": 0.7182841708896324, + "learning_rate": 2.6160185803858926e-07, + "loss": 0.1705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09732286632061005, + "step": 12110, + "valid_targets_mean": 3232.2, + "valid_targets_min": 1599 + }, + { + "epoch": 6.678610804851157, + "grad_norm": 0.6819128107361788, + "learning_rate": 2.5718875541820285e-07, + "loss": 0.1676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08373957127332687, + "step": 12115, + "valid_targets_mean": 3505.6, + "valid_targets_min": 776 + }, + { + "epoch": 6.681367144432194, + "grad_norm": 0.6567758874646434, + "learning_rate": 2.5281295274489413e-07, + "loss": 0.1678, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10158775746822357, + "step": 12120, + "valid_targets_mean": 4465.2, + "valid_targets_min": 1866 + }, + { + "epoch": 6.684123484013231, + "grad_norm": 0.638350300715274, + "learning_rate": 2.484744582858389e-07, + "loss": 0.1593, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08434487879276276, + "step": 12125, + "valid_targets_mean": 3549.2, + "valid_targets_min": 735 + }, + { + "epoch": 6.686879823594267, + "grad_norm": 0.6370678710667508, + "learning_rate": 2.4417328023771835e-07, + "loss": 0.1625, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0740165114402771, + "step": 12130, + "valid_targets_mean": 2900.2, + "valid_targets_min": 1470 + }, + { + "epoch": 6.689636163175303, + "grad_norm": 0.6906262698258795, + "learning_rate": 2.399094267267188e-07, + "loss": 0.1685, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06469167023897171, + "step": 12135, + "valid_targets_mean": 2116.2, + "valid_targets_min": 373 + }, + { + "epoch": 6.69239250275634, + "grad_norm": 0.7356953381100514, + "learning_rate": 2.3568290580850307e-07, + "loss": 0.1817, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07435179501771927, + "step": 12140, + "valid_targets_mean": 2585.5, + "valid_targets_min": 952 + }, + { + "epoch": 6.695148842337376, + "grad_norm": 0.6997449466562002, + "learning_rate": 2.3149372546821258e-07, + "loss": 0.1629, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07441840320825577, + "step": 12145, + "valid_targets_mean": 2983.9, + "valid_targets_min": 1505 + }, + { + "epoch": 6.697905181918412, + "grad_norm": 0.6495561502209735, + "learning_rate": 2.273418936204297e-07, + "loss": 0.166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07181662321090698, + "step": 12150, + "valid_targets_mean": 3333.4, + "valid_targets_min": 633 + }, + { + "epoch": 6.700661521499448, + "grad_norm": 0.6619915341560292, + "learning_rate": 2.2322741810918202e-07, + "loss": 0.1622, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09099540114402771, + "step": 12155, + "valid_targets_mean": 3836.8, + "valid_targets_min": 1483 + }, + { + "epoch": 6.703417861080485, + "grad_norm": 0.6170446651285765, + "learning_rate": 2.1915030670791594e-07, + "loss": 0.1701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07517152279615402, + "step": 12160, + "valid_targets_mean": 3521.6, + "valid_targets_min": 1797 + }, + { + "epoch": 6.706174200661522, + "grad_norm": 0.6983795483191304, + "learning_rate": 2.1511056711948751e-07, + "loss": 0.1775, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08942985534667969, + "step": 12165, + "valid_targets_mean": 3853.1, + "valid_targets_min": 2662 + }, + { + "epoch": 6.708930540242558, + "grad_norm": 0.6502312090824374, + "learning_rate": 2.1110820697614943e-07, + "loss": 0.1748, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09085971862077713, + "step": 12170, + "valid_targets_mean": 4088.1, + "valid_targets_min": 2230 + }, + { + "epoch": 6.711686879823594, + "grad_norm": 0.6556161982486461, + "learning_rate": 2.071432338395285e-07, + "loss": 0.1671, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07092592120170593, + "step": 12175, + "valid_targets_mean": 3464.5, + "valid_targets_min": 1628 + }, + { + "epoch": 6.714443219404631, + "grad_norm": 0.68528133779671, + "learning_rate": 2.0321565520061925e-07, + "loss": 0.1694, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11031152307987213, + "step": 12180, + "valid_targets_mean": 4436.0, + "valid_targets_min": 830 + }, + { + "epoch": 6.7171995589856675, + "grad_norm": 0.5715519989981075, + "learning_rate": 1.9932547847976823e-07, + "loss": 0.1671, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0657581239938736, + "step": 12185, + "valid_targets_mean": 3477.5, + "valid_targets_min": 1715 + }, + { + "epoch": 6.719955898566703, + "grad_norm": 0.4965272792620506, + "learning_rate": 1.9547271102665854e-07, + "loss": 0.1368, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06445446610450745, + "step": 12190, + "valid_targets_mean": 2246.1, + "valid_targets_min": 331 + }, + { + "epoch": 6.722712238147739, + "grad_norm": 0.39159309745209836, + "learning_rate": 1.91657360120292e-07, + "loss": 0.0831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03702632337808609, + "step": 12195, + "valid_targets_mean": 3604.4, + "valid_targets_min": 708 + }, + { + "epoch": 6.725468577728776, + "grad_norm": 0.4488497735740002, + "learning_rate": 1.878794329689848e-07, + "loss": 0.074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03925613313913345, + "step": 12200, + "valid_targets_mean": 2933.4, + "valid_targets_min": 729 + }, + { + "epoch": 6.728224917309813, + "grad_norm": 0.49531909474119057, + "learning_rate": 1.8413893671034744e-07, + "loss": 0.0793, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07102900743484497, + "step": 12205, + "valid_targets_mean": 2718.9, + "valid_targets_min": 921 + }, + { + "epoch": 6.730981256890849, + "grad_norm": 0.3644781154715217, + "learning_rate": 1.804358784112692e-07, + "loss": 0.07, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02908411994576454, + "step": 12210, + "valid_targets_mean": 2827.9, + "valid_targets_min": 813 + }, + { + "epoch": 6.733737596471885, + "grad_norm": 0.4629293801081908, + "learning_rate": 1.7677026506791372e-07, + "loss": 0.0774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0412607416510582, + "step": 12215, + "valid_targets_mean": 2883.1, + "valid_targets_min": 854 + }, + { + "epoch": 6.736493936052922, + "grad_norm": 0.7874278638976012, + "learning_rate": 1.731421036056946e-07, + "loss": 0.1174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04758979380130768, + "step": 12220, + "valid_targets_mean": 1034.4, + "valid_targets_min": 536 + }, + { + "epoch": 6.7392502756339585, + "grad_norm": 0.43685330684150553, + "learning_rate": 1.6955140087927314e-07, + "loss": 0.0881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035918593406677246, + "step": 12225, + "valid_targets_mean": 2992.8, + "valid_targets_min": 915 + }, + { + "epoch": 6.7420066152149944, + "grad_norm": 0.4641480633861165, + "learning_rate": 1.6599816367253385e-07, + "loss": 0.0832, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04227231815457344, + "step": 12230, + "valid_targets_mean": 3356.0, + "valid_targets_min": 2281 + }, + { + "epoch": 6.74476295479603, + "grad_norm": 0.38070374876702257, + "learning_rate": 1.6248239869858463e-07, + "loss": 0.1087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028317920863628387, + "step": 12235, + "valid_targets_mean": 3776.1, + "valid_targets_min": 958 + }, + { + "epoch": 6.747519294377067, + "grad_norm": 0.41804036329646377, + "learning_rate": 1.5900411259973214e-07, + "loss": 0.0634, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04160662367939949, + "step": 12240, + "valid_targets_mean": 4223.1, + "valid_targets_min": 2082 + }, + { + "epoch": 6.750275633958104, + "grad_norm": 0.4365525412249055, + "learning_rate": 1.5556331194747755e-07, + "loss": 0.0732, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04053156077861786, + "step": 12245, + "valid_targets_mean": 4378.9, + "valid_targets_min": 3337 + }, + { + "epoch": 6.75303197353914, + "grad_norm": 0.4044760928443498, + "learning_rate": 1.5216000324250079e-07, + "loss": 0.0728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034392617642879486, + "step": 12250, + "valid_targets_mean": 3910.5, + "valid_targets_min": 515 + }, + { + "epoch": 6.755788313120176, + "grad_norm": 0.5947938336864041, + "learning_rate": 1.4879419291464747e-07, + "loss": 0.1143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04072687774896622, + "step": 12255, + "valid_targets_mean": 1878.6, + "valid_targets_min": 546 + }, + { + "epoch": 6.758544652701213, + "grad_norm": 0.3885159988411775, + "learning_rate": 1.4546588732291532e-07, + "loss": 0.067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0334409698843956, + "step": 12260, + "valid_targets_mean": 3397.5, + "valid_targets_min": 2715 + }, + { + "epoch": 6.76130099228225, + "grad_norm": 0.5414121397679493, + "learning_rate": 1.4217509275545217e-07, + "loss": 0.0751, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03250187635421753, + "step": 12265, + "valid_targets_mean": 1612.8, + "valid_targets_min": 515 + }, + { + "epoch": 6.7640573318632855, + "grad_norm": 0.3728282973445053, + "learning_rate": 1.3892181542952688e-07, + "loss": 0.0626, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02214132994413376, + "step": 12270, + "valid_targets_mean": 3322.8, + "valid_targets_min": 840 + }, + { + "epoch": 6.766813671444321, + "grad_norm": 0.487910828485507, + "learning_rate": 1.357060614915362e-07, + "loss": 0.0766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03644410893321037, + "step": 12275, + "valid_targets_mean": 2764.0, + "valid_targets_min": 574 + }, + { + "epoch": 6.769570011025358, + "grad_norm": 0.4787781616930142, + "learning_rate": 1.3252783701698026e-07, + "loss": 0.0783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040074825286865234, + "step": 12280, + "valid_targets_mean": 2077.9, + "valid_targets_min": 847 + }, + { + "epoch": 6.772326350606395, + "grad_norm": 0.4688174219240074, + "learning_rate": 1.2938714801044916e-07, + "loss": 0.084, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04233285412192345, + "step": 12285, + "valid_targets_mean": 2857.1, + "valid_targets_min": 770 + }, + { + "epoch": 6.775082690187431, + "grad_norm": 0.42421671162481395, + "learning_rate": 1.2628400040562983e-07, + "loss": 0.0814, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03407226875424385, + "step": 12290, + "valid_targets_mean": 1736.5, + "valid_targets_min": 611 + }, + { + "epoch": 6.777839029768468, + "grad_norm": 0.5381281433555622, + "learning_rate": 1.2321840006527475e-07, + "loss": 0.1097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05816757678985596, + "step": 12295, + "valid_targets_mean": 3113.2, + "valid_targets_min": 1665 + }, + { + "epoch": 6.780595369349504, + "grad_norm": 0.32634658730962146, + "learning_rate": 1.2019035278119983e-07, + "loss": 0.0699, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028720654547214508, + "step": 12300, + "valid_targets_mean": 4061.4, + "valid_targets_min": 3722 + }, + { + "epoch": 6.783351708930541, + "grad_norm": 0.4402699596600852, + "learning_rate": 1.1719986427427111e-07, + "loss": 0.0774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035613298416137695, + "step": 12305, + "valid_targets_mean": 1199.0, + "valid_targets_min": 711 + }, + { + "epoch": 6.786108048511577, + "grad_norm": 0.4221270971725821, + "learning_rate": 1.1424694019439575e-07, + "loss": 0.1031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035849615931510925, + "step": 12310, + "valid_targets_mean": 3407.8, + "valid_targets_min": 1803 + }, + { + "epoch": 6.788864388092613, + "grad_norm": 0.4025528661508341, + "learning_rate": 1.1133158612051331e-07, + "loss": 0.0894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03580104187130928, + "step": 12315, + "valid_targets_mean": 3789.4, + "valid_targets_min": 3167 + }, + { + "epoch": 6.791620727673649, + "grad_norm": 0.7328840761526654, + "learning_rate": 1.084538075605801e-07, + "loss": 0.099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09247465431690216, + "step": 12320, + "valid_targets_mean": 1659.2, + "valid_targets_min": 864 + }, + { + "epoch": 6.794377067254686, + "grad_norm": 0.47053493384841494, + "learning_rate": 1.0561360995156034e-07, + "loss": 0.2354, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0420159250497818, + "step": 12325, + "valid_targets_mean": 3663.0, + "valid_targets_min": 2876 + }, + { + "epoch": 6.797133406835722, + "grad_norm": 0.4577131535691857, + "learning_rate": 1.0281099865941946e-07, + "loss": 0.0844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03703731298446655, + "step": 12330, + "valid_targets_mean": 3140.6, + "valid_targets_min": 828 + }, + { + "epoch": 6.799889746416759, + "grad_norm": 0.552584779678047, + "learning_rate": 1.0004597897911084e-07, + "loss": 0.0831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060297369956970215, + "step": 12335, + "valid_targets_mean": 3309.0, + "valid_targets_min": 1794 + }, + { + "epoch": 6.802646085997795, + "grad_norm": 0.45787766866796586, + "learning_rate": 9.731855613456687e-08, + "loss": 0.1682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04826197028160095, + "step": 12340, + "valid_targets_mean": 3527.5, + "valid_targets_min": 2771 + }, + { + "epoch": 6.805402425578832, + "grad_norm": 0.3464777202015583, + "learning_rate": 9.462873527868343e-08, + "loss": 0.0633, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029064536094665527, + "step": 12345, + "valid_targets_mean": 4565.0, + "valid_targets_min": 2655 + }, + { + "epoch": 6.808158765159868, + "grad_norm": 0.38391979139373805, + "learning_rate": 9.197652149332215e-08, + "loss": 0.0847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.020441971719264984, + "step": 12350, + "valid_targets_mean": 2743.9, + "valid_targets_min": 557 + }, + { + "epoch": 6.8109151047409044, + "grad_norm": 0.43716904533030027, + "learning_rate": 8.936191978929254e-08, + "loss": 0.0818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03986711800098419, + "step": 12355, + "valid_targets_mean": 4249.6, + "valid_targets_min": 1606 + }, + { + "epoch": 6.81367144432194, + "grad_norm": 0.46223202661089224, + "learning_rate": 8.678493510633656e-08, + "loss": 0.0692, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03258770704269409, + "step": 12360, + "valid_targets_mean": 3533.0, + "valid_targets_min": 2163 + }, + { + "epoch": 6.816427783902977, + "grad_norm": 0.45594344277095267, + "learning_rate": 8.424557231313746e-08, + "loss": 0.0679, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03353439271450043, + "step": 12365, + "valid_targets_mean": 3057.5, + "valid_targets_min": 533 + }, + { + "epoch": 6.819184123484013, + "grad_norm": 0.46276468353156036, + "learning_rate": 8.174383620729087e-08, + "loss": 0.0699, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034651368856430054, + "step": 12370, + "valid_targets_mean": 3622.0, + "valid_targets_min": 2249 + }, + { + "epoch": 6.82194046306505, + "grad_norm": 0.393765785510414, + "learning_rate": 7.927973151531155e-08, + "loss": 0.0668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02829364314675331, + "step": 12375, + "valid_targets_mean": 2801.5, + "valid_targets_min": 817 + }, + { + "epoch": 6.824696802646086, + "grad_norm": 0.697610170078838, + "learning_rate": 7.685326289261108e-08, + "loss": 0.1063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09946341812610626, + "step": 12380, + "valid_targets_mean": 1954.6, + "valid_targets_min": 859 + }, + { + "epoch": 6.827453142227123, + "grad_norm": 0.3943799984537571, + "learning_rate": 7.446443492350242e-08, + "loss": 0.0643, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03494356572628021, + "step": 12385, + "valid_targets_mean": 3564.0, + "valid_targets_min": 2709 + }, + { + "epoch": 6.830209481808159, + "grad_norm": 0.6391541782935491, + "learning_rate": 7.211325212117759e-08, + "loss": 0.0634, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04789785295724869, + "step": 12390, + "valid_targets_mean": 1605.5, + "valid_targets_min": 781 + }, + { + "epoch": 6.8329658213891955, + "grad_norm": 0.3509129194686773, + "learning_rate": 6.979971892770998e-08, + "loss": 0.0637, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026542456820607185, + "step": 12395, + "valid_targets_mean": 3645.4, + "valid_targets_min": 2141 + }, + { + "epoch": 6.835722160970231, + "grad_norm": 0.4007563669851462, + "learning_rate": 6.752383971404097e-08, + "loss": 0.0648, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02687971293926239, + "step": 12400, + "valid_targets_mean": 1990.2, + "valid_targets_min": 545 + }, + { + "epoch": 6.838478500551268, + "grad_norm": 0.4573670857530484, + "learning_rate": 6.528561877997108e-08, + "loss": 0.1049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0337439589202404, + "step": 12405, + "valid_targets_mean": 3322.1, + "valid_targets_min": 904 + }, + { + "epoch": 6.841234840132304, + "grad_norm": 0.42799170664414515, + "learning_rate": 6.308506035415552e-08, + "loss": 0.0923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06767337769269943, + "step": 12410, + "valid_targets_mean": 3514.5, + "valid_targets_min": 2184 + }, + { + "epoch": 6.843991179713341, + "grad_norm": 0.3547379707103964, + "learning_rate": 6.092216859409306e-08, + "loss": 0.0776, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.022518187761306763, + "step": 12415, + "valid_targets_mean": 1680.6, + "valid_targets_min": 595 + }, + { + "epoch": 6.846747519294377, + "grad_norm": 0.9754049777353883, + "learning_rate": 5.879694758611942e-08, + "loss": 0.0911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0792594701051712, + "step": 12420, + "valid_targets_mean": 1323.4, + "valid_targets_min": 575 + }, + { + "epoch": 6.849503858875414, + "grad_norm": 0.6090915378505878, + "learning_rate": 5.6709401345396154e-08, + "loss": 0.1138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04375186935067177, + "step": 12425, + "valid_targets_mean": 1671.9, + "valid_targets_min": 691 + }, + { + "epoch": 6.85226019845645, + "grad_norm": 0.49608920299331516, + "learning_rate": 5.4659533815912826e-08, + "loss": 0.0794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045576900243759155, + "step": 12430, + "valid_targets_mean": 1879.5, + "valid_targets_min": 687 + }, + { + "epoch": 6.855016538037487, + "grad_norm": 0.3996704633250836, + "learning_rate": 5.264734887046707e-08, + "loss": 0.0929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03048473782837391, + "step": 12435, + "valid_targets_mean": 3679.5, + "valid_targets_min": 2015 + }, + { + "epoch": 6.8577728776185225, + "grad_norm": 0.6797079861417227, + "learning_rate": 5.06728503106646e-08, + "loss": 0.0744, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04055529087781906, + "step": 12440, + "valid_targets_mean": 1331.5, + "valid_targets_min": 520 + }, + { + "epoch": 6.860529217199559, + "grad_norm": 0.3479325288403404, + "learning_rate": 4.8736041866912496e-08, + "loss": 0.0725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027241922914981842, + "step": 12445, + "valid_targets_mean": 4111.8, + "valid_targets_min": 466 + }, + { + "epoch": 6.863285556780595, + "grad_norm": 0.30301669227404565, + "learning_rate": 4.683692719840594e-08, + "loss": 0.0782, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0210411474108696, + "step": 12450, + "valid_targets_mean": 5544.5, + "valid_targets_min": 4219 + }, + { + "epoch": 6.866041896361632, + "grad_norm": 0.36646571380041976, + "learning_rate": 4.497550989313038e-08, + "loss": 0.068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02499362640082836, + "step": 12455, + "valid_targets_mean": 3800.1, + "valid_targets_min": 456 + }, + { + "epoch": 6.868798235942668, + "grad_norm": 0.4231477768867857, + "learning_rate": 4.315179346784826e-08, + "loss": 0.0663, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031195340678095818, + "step": 12460, + "valid_targets_mean": 3487.1, + "valid_targets_min": 976 + }, + { + "epoch": 6.871554575523705, + "grad_norm": 0.5395522968770873, + "learning_rate": 4.136578136809233e-08, + "loss": 0.0756, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046962156891822815, + "step": 12465, + "valid_targets_mean": 2431.6, + "valid_targets_min": 591 + }, + { + "epoch": 6.874310915104741, + "grad_norm": 0.42043925972427054, + "learning_rate": 3.96174769681612e-08, + "loss": 0.0802, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0466318242251873, + "step": 12470, + "valid_targets_mean": 3660.8, + "valid_targets_min": 2753 + }, + { + "epoch": 6.877067254685778, + "grad_norm": 0.43272635189029723, + "learning_rate": 3.790688357111494e-08, + "loss": 0.0848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0430942177772522, + "step": 12475, + "valid_targets_mean": 3661.6, + "valid_targets_min": 2892 + }, + { + "epoch": 6.879823594266814, + "grad_norm": 0.3709117303350464, + "learning_rate": 3.62340044087639e-08, + "loss": 0.0725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03121134452521801, + "step": 12480, + "valid_targets_mean": 3447.0, + "valid_targets_min": 1093 + }, + { + "epoch": 6.88257993384785, + "grad_norm": 0.44023685294613685, + "learning_rate": 3.45988426416688e-08, + "loss": 0.0712, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044934310019016266, + "step": 12485, + "valid_targets_mean": 3798.8, + "valid_targets_min": 1861 + }, + { + "epoch": 6.885336273428886, + "grad_norm": 0.39168424138780067, + "learning_rate": 3.3001401359125106e-08, + "loss": 0.0799, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02861708588898182, + "step": 12490, + "valid_targets_mean": 3005.0, + "valid_targets_min": 720 + }, + { + "epoch": 6.888092613009923, + "grad_norm": 0.37415787807270384, + "learning_rate": 3.1441683579167545e-08, + "loss": 0.0586, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02680686116218567, + "step": 12495, + "valid_targets_mean": 3779.8, + "valid_targets_min": 2761 + }, + { + "epoch": 6.890848952590959, + "grad_norm": 0.46973470535769046, + "learning_rate": 2.991969224856117e-08, + "loss": 0.0796, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030520513653755188, + "step": 12500, + "valid_targets_mean": 2407.5, + "valid_targets_min": 588 + }, + { + "epoch": 6.893605292171996, + "grad_norm": 0.41765001307274974, + "learning_rate": 2.8435430242792496e-08, + "loss": 0.0739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03649859130382538, + "step": 12505, + "valid_targets_mean": 3408.8, + "valid_targets_min": 596 + }, + { + "epoch": 6.896361631753032, + "grad_norm": 0.44005214520196234, + "learning_rate": 2.6988900366067274e-08, + "loss": 0.0832, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04240091145038605, + "step": 12510, + "valid_targets_mean": 3530.0, + "valid_targets_min": 1636 + }, + { + "epoch": 6.899117971334069, + "grad_norm": 0.5143396952701793, + "learning_rate": 2.558010535130162e-08, + "loss": 0.0762, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03768397495150566, + "step": 12515, + "valid_targets_mean": 1950.8, + "valid_targets_min": 454 + }, + { + "epoch": 6.901874310915105, + "grad_norm": 0.785720658083393, + "learning_rate": 2.4209047860122015e-08, + "loss": 0.0834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04537493735551834, + "step": 12520, + "valid_targets_mean": 819.5, + "valid_targets_min": 609 + }, + { + "epoch": 6.904630650496141, + "grad_norm": 0.5631617906200488, + "learning_rate": 2.2875730482856405e-08, + "loss": 0.0779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06652190536260605, + "step": 12525, + "valid_targets_mean": 2364.8, + "valid_targets_min": 893 + }, + { + "epoch": 6.907386990077177, + "grad_norm": 0.4190874090273679, + "learning_rate": 2.1580155738534226e-08, + "loss": 0.0834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030893828719854355, + "step": 12530, + "valid_targets_mean": 3029.5, + "valid_targets_min": 1140 + }, + { + "epoch": 6.910143329658214, + "grad_norm": 0.4623244041410091, + "learning_rate": 2.0322326074875277e-08, + "loss": 0.076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04606953263282776, + "step": 12535, + "valid_targets_mean": 3029.6, + "valid_targets_min": 1502 + }, + { + "epoch": 6.91289966923925, + "grad_norm": 0.6037040463989813, + "learning_rate": 1.9102243868285298e-08, + "loss": 0.0686, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042129166424274445, + "step": 12540, + "valid_targets_mean": 1295.8, + "valid_targets_min": 624 + }, + { + "epoch": 6.915656008820287, + "grad_norm": 0.6057010111997407, + "learning_rate": 1.791991142385596e-08, + "loss": 0.0849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0490148663520813, + "step": 12545, + "valid_targets_mean": 1481.6, + "valid_targets_min": 687 + }, + { + "epoch": 6.918412348401323, + "grad_norm": 0.3527213672917072, + "learning_rate": 1.677533097536266e-08, + "loss": 0.0731, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02344234474003315, + "step": 12550, + "valid_targets_mean": 3963.5, + "valid_targets_min": 3651 + }, + { + "epoch": 6.92116868798236, + "grad_norm": 0.3901533639568866, + "learning_rate": 1.5668504685251163e-08, + "loss": 0.0653, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04337751492857933, + "step": 12555, + "valid_targets_mean": 5041.0, + "valid_targets_min": 3813 + }, + { + "epoch": 6.923925027563396, + "grad_norm": 0.3760187874521229, + "learning_rate": 1.4599434644639866e-08, + "loss": 0.0732, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034422390162944794, + "step": 12560, + "valid_targets_mean": 2392.4, + "valid_targets_min": 757 + }, + { + "epoch": 6.9266813671444325, + "grad_norm": 0.39913857425325283, + "learning_rate": 1.3568122873313106e-08, + "loss": 0.0641, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035604096949100494, + "step": 12565, + "valid_targets_mean": 3566.1, + "valid_targets_min": 664 + }, + { + "epoch": 6.929437706725468, + "grad_norm": 0.4194202853594675, + "learning_rate": 1.2574571319723395e-08, + "loss": 0.0628, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04720073938369751, + "step": 12570, + "valid_targets_mean": 4990.0, + "valid_targets_min": 4088 + }, + { + "epoch": 6.932194046306505, + "grad_norm": 0.3769332722869115, + "learning_rate": 1.161878186098031e-08, + "loss": 0.081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0317886620759964, + "step": 12575, + "valid_targets_mean": 3220.0, + "valid_targets_min": 779 + }, + { + "epoch": 6.934950385887541, + "grad_norm": 0.5484818281743982, + "learning_rate": 1.070075630284828e-08, + "loss": 0.0896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03956032171845436, + "step": 12580, + "valid_targets_mean": 3469.2, + "valid_targets_min": 1084 + }, + { + "epoch": 6.937706725468578, + "grad_norm": 0.4329248881170913, + "learning_rate": 9.820496379748801e-09, + "loss": 0.0727, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0434672087430954, + "step": 12585, + "valid_targets_mean": 3214.4, + "valid_targets_min": 700 + }, + { + "epoch": 6.940463065049614, + "grad_norm": 0.37173091753960186, + "learning_rate": 8.978003754744891e-09, + "loss": 0.0757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02790200337767601, + "step": 12590, + "valid_targets_mean": 2815.9, + "valid_targets_min": 769 + }, + { + "epoch": 6.943219404630651, + "grad_norm": 0.34561512876224665, + "learning_rate": 8.17328001955886e-09, + "loss": 0.0926, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027209173887968063, + "step": 12595, + "valid_targets_mean": 3541.8, + "valid_targets_min": 982 + }, + { + "epoch": 6.945975744211687, + "grad_norm": 0.4606705745452347, + "learning_rate": 7.406326694543442e-09, + "loss": 0.07, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04213263839483261, + "step": 12600, + "valid_targets_mean": 3189.4, + "valid_targets_min": 1043 + }, + { + "epoch": 6.948732083792724, + "grad_norm": 0.45090575906281133, + "learning_rate": 6.677145228699555e-09, + "loss": 0.0743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033880386501550674, + "step": 12605, + "valid_targets_mean": 3380.1, + "valid_targets_min": 1067 + }, + { + "epoch": 6.9514884233737595, + "grad_norm": 0.41481289839703306, + "learning_rate": 5.985736999665204e-09, + "loss": 0.0709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0549340583384037, + "step": 12610, + "valid_targets_mean": 3086.6, + "valid_targets_min": 901 + }, + { + "epoch": 6.954244762954796, + "grad_norm": 0.5924465611837193, + "learning_rate": 5.332103313713255e-09, + "loss": 0.105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047529011964797974, + "step": 12615, + "valid_targets_mean": 1287.2, + "valid_targets_min": 567 + }, + { + "epoch": 6.957001102535832, + "grad_norm": 0.4657899649635288, + "learning_rate": 4.716245405747e-09, + "loss": 0.0827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048263490200042725, + "step": 12620, + "valid_targets_mean": 3460.5, + "valid_targets_min": 831 + }, + { + "epoch": 6.959757442116869, + "grad_norm": 0.43169108751587437, + "learning_rate": 4.138164439304593e-09, + "loss": 0.0702, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029419872909784317, + "step": 12625, + "valid_targets_mean": 2148.4, + "valid_targets_min": 632 + }, + { + "epoch": 6.962513781697905, + "grad_norm": 0.37940664316801204, + "learning_rate": 3.5978615065479504e-09, + "loss": 0.0699, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034746430814266205, + "step": 12630, + "valid_targets_mean": 2711.9, + "valid_targets_min": 1583 + }, + { + "epoch": 6.965270121278942, + "grad_norm": 0.3871602342661975, + "learning_rate": 3.0953376282694126e-09, + "loss": 0.0686, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029064588248729706, + "step": 12635, + "valid_targets_mean": 3267.4, + "valid_targets_min": 796 + }, + { + "epoch": 6.968026460859978, + "grad_norm": 0.4060599187356548, + "learning_rate": 2.6305937538828595e-09, + "loss": 0.0622, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029700959101319313, + "step": 12640, + "valid_targets_mean": 3208.9, + "valid_targets_min": 952 + }, + { + "epoch": 6.970782800441015, + "grad_norm": 0.5460301046026552, + "learning_rate": 2.203630761423714e-09, + "loss": 0.0861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03830365464091301, + "step": 12645, + "valid_targets_mean": 2115.5, + "valid_targets_min": 885 + }, + { + "epoch": 6.9735391400220506, + "grad_norm": 0.4204527355365197, + "learning_rate": 1.81444945755338e-09, + "loss": 0.0626, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03788033127784729, + "step": 12650, + "valid_targets_mean": 3298.5, + "valid_targets_min": 1018 + }, + { + "epoch": 6.976295479603087, + "grad_norm": 0.4823043022092928, + "learning_rate": 1.4630505775459213e-09, + "loss": 0.068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03535851836204529, + "step": 12655, + "valid_targets_mean": 2170.4, + "valid_targets_min": 782 + }, + { + "epoch": 6.979051819184123, + "grad_norm": 0.4590508677214216, + "learning_rate": 1.1494347852991639e-09, + "loss": 0.069, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04394698143005371, + "step": 12660, + "valid_targets_mean": 3039.9, + "valid_targets_min": 2543 + }, + { + "epoch": 6.98180815876516, + "grad_norm": 0.3463589150194841, + "learning_rate": 8.736026733213721e-10, + "loss": 0.0716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03467486798763275, + "step": 12665, + "valid_targets_mean": 3951.6, + "valid_targets_min": 2264 + }, + { + "epoch": 6.984564498346196, + "grad_norm": 0.43211476932401066, + "learning_rate": 6.355547627445724e-10, + "loss": 0.0615, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041309576481580734, + "step": 12670, + "valid_targets_mean": 4005.2, + "valid_targets_min": 2423 + }, + { + "epoch": 6.987320837927233, + "grad_norm": 0.3974107033075562, + "learning_rate": 4.352915033090099e-10, + "loss": 0.0674, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043060604482889175, + "step": 12675, + "valid_targets_mean": 4350.4, + "valid_targets_min": 2545 + }, + { + "epoch": 6.990077177508269, + "grad_norm": 0.5617988282988646, + "learning_rate": 2.728132733720301e-10, + "loss": 0.1008, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07315270602703094, + "step": 12680, + "valid_targets_mean": 2167.4, + "valid_targets_min": 1523 + }, + { + "epoch": 6.992833517089306, + "grad_norm": 0.3364096978609004, + "learning_rate": 1.4812037989919702e-10, + "loss": 0.1177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.020261526107788086, + "step": 12685, + "valid_targets_mean": 3601.6, + "valid_targets_min": 1092 + }, + { + "epoch": 6.995589856670342, + "grad_norm": 0.3638862603136978, + "learning_rate": 6.121305847317516e-11, + "loss": 0.0691, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026766205206513405, + "step": 12690, + "valid_targets_mean": 3158.4, + "valid_targets_min": 835 + }, + { + "epoch": 6.998346196251378, + "grad_norm": 0.3447680042561147, + "learning_rate": 1.2091473289288502e-11, + "loss": 0.0655, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030071251094341278, + "step": 12695, + "valid_targets_mean": 3701.5, + "valid_targets_min": 2543 + }, + { + "epoch": 7.0, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042520757764577866, + "step": 12698, + "total_flos": 3.824454126066991e+18, + "train_loss": 0.1807381806187807, + "train_runtime": 81415.8671, + "train_samples_per_second": 2.495, + "train_steps_per_second": 0.156, + "valid_targets_mean": 4458.2, + "valid_targets_min": 1268 + } + ], + "logging_steps": 5, + "max_steps": 12698, + "num_input_tokens_seen": 0, + "num_train_epochs": 7, + "save_steps": 1500, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": true + }, + "attributes": {} + } + }, + "total_flos": 3.824454126066991e+18, + "train_batch_size": 1, + "trial_name": null, + "trial_params": null +}