64_Bash_GPT5_Traces / trainer_state.json
Jashan887's picture
Upload folder using huggingface_hub
ba08196 verified
{
"best_global_step": null,
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 6.0,
"eval_steps": 500,
"global_step": 2394,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.012531328320802004,
"grad_norm": 11.967602793791775,
"learning_rate": 6.666666666666667e-07,
"loss": 0.9827,
"loss_nan_ranks": 0,
"loss_rank_avg": 1.0594820976257324,
"step": 5,
"valid_targets_mean": 1399.4,
"valid_targets_min": 664
},
{
"epoch": 0.02506265664160401,
"grad_norm": 8.98761722777954,
"learning_rate": 1.5e-06,
"loss": 0.943,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.920002818107605,
"step": 10,
"valid_targets_mean": 1760.4,
"valid_targets_min": 580
},
{
"epoch": 0.03759398496240601,
"grad_norm": 7.940426239884903,
"learning_rate": 2.3333333333333336e-06,
"loss": 0.9234,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.9283522963523865,
"step": 15,
"valid_targets_mean": 1458.5,
"valid_targets_min": 677
},
{
"epoch": 0.05012531328320802,
"grad_norm": 5.224005967961386,
"learning_rate": 3.1666666666666667e-06,
"loss": 0.884,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.8793013095855713,
"step": 20,
"valid_targets_mean": 1278.1,
"valid_targets_min": 680
},
{
"epoch": 0.06265664160401002,
"grad_norm": 3.7348565637539166,
"learning_rate": 4.000000000000001e-06,
"loss": 0.8446,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.8814775943756104,
"step": 25,
"valid_targets_mean": 1153.8,
"valid_targets_min": 728
},
{
"epoch": 0.07518796992481203,
"grad_norm": 2.075582535169858,
"learning_rate": 4.833333333333333e-06,
"loss": 0.7618,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.6769101619720459,
"step": 30,
"valid_targets_mean": 1775.1,
"valid_targets_min": 597
},
{
"epoch": 0.08771929824561403,
"grad_norm": 1.651201460536687,
"learning_rate": 5.666666666666667e-06,
"loss": 0.7189,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.7634528279304504,
"step": 35,
"valid_targets_mean": 1523.2,
"valid_targets_min": 699
},
{
"epoch": 0.10025062656641603,
"grad_norm": 1.5275695524844886,
"learning_rate": 6.5000000000000004e-06,
"loss": 0.7277,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.750813364982605,
"step": 40,
"valid_targets_mean": 1344.6,
"valid_targets_min": 679
},
{
"epoch": 0.11278195488721804,
"grad_norm": 1.361371709568494,
"learning_rate": 7.333333333333333e-06,
"loss": 0.703,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.6691080927848816,
"step": 45,
"valid_targets_mean": 1303.0,
"valid_targets_min": 734
},
{
"epoch": 0.12531328320802004,
"grad_norm": 1.152708781810184,
"learning_rate": 8.166666666666668e-06,
"loss": 0.6218,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.6578604578971863,
"step": 50,
"valid_targets_mean": 1385.8,
"valid_targets_min": 715
},
{
"epoch": 0.13784461152882205,
"grad_norm": 0.9820138424429671,
"learning_rate": 9e-06,
"loss": 0.6597,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.6535561084747314,
"step": 55,
"valid_targets_mean": 1496.4,
"valid_targets_min": 717
},
{
"epoch": 0.15037593984962405,
"grad_norm": 1.158994935896899,
"learning_rate": 9.833333333333333e-06,
"loss": 0.6282,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.7068292498588562,
"step": 60,
"valid_targets_mean": 1289.7,
"valid_targets_min": 666
},
{
"epoch": 0.16290726817042606,
"grad_norm": 0.9426226170878005,
"learning_rate": 1.0666666666666667e-05,
"loss": 0.5982,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.585883378982544,
"step": 65,
"valid_targets_mean": 1464.9,
"valid_targets_min": 601
},
{
"epoch": 0.17543859649122806,
"grad_norm": 0.9405312188936708,
"learning_rate": 1.15e-05,
"loss": 0.5808,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.6200866103172302,
"step": 70,
"valid_targets_mean": 1546.4,
"valid_targets_min": 667
},
{
"epoch": 0.18796992481203006,
"grad_norm": 1.0123323450686448,
"learning_rate": 1.2333333333333334e-05,
"loss": 0.5854,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.6166951060295105,
"step": 75,
"valid_targets_mean": 1234.2,
"valid_targets_min": 603
},
{
"epoch": 0.20050125313283207,
"grad_norm": 0.994487225478387,
"learning_rate": 1.3166666666666667e-05,
"loss": 0.5918,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5867680311203003,
"step": 80,
"valid_targets_mean": 1479.2,
"valid_targets_min": 654
},
{
"epoch": 0.21303258145363407,
"grad_norm": 0.8431318418016086,
"learning_rate": 1.4e-05,
"loss": 0.5669,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5588182210922241,
"step": 85,
"valid_targets_mean": 1593.0,
"valid_targets_min": 641
},
{
"epoch": 0.22556390977443608,
"grad_norm": 0.9663714100983305,
"learning_rate": 1.4833333333333336e-05,
"loss": 0.5683,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.601900041103363,
"step": 90,
"valid_targets_mean": 1452.9,
"valid_targets_min": 647
},
{
"epoch": 0.23809523809523808,
"grad_norm": 0.9262179876650385,
"learning_rate": 1.5666666666666667e-05,
"loss": 0.5344,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5362147092819214,
"step": 95,
"valid_targets_mean": 1426.3,
"valid_targets_min": 705
},
{
"epoch": 0.2506265664160401,
"grad_norm": 0.8714735709336389,
"learning_rate": 1.65e-05,
"loss": 0.5704,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5262184739112854,
"step": 100,
"valid_targets_mean": 1516.5,
"valid_targets_min": 683
},
{
"epoch": 0.2631578947368421,
"grad_norm": 0.8918423929105987,
"learning_rate": 1.7333333333333336e-05,
"loss": 0.5614,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5865960717201233,
"step": 105,
"valid_targets_mean": 1535.5,
"valid_targets_min": 671
},
{
"epoch": 0.2756892230576441,
"grad_norm": 0.8935023571966156,
"learning_rate": 1.8166666666666667e-05,
"loss": 0.538,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5343993306159973,
"step": 110,
"valid_targets_mean": 1453.6,
"valid_targets_min": 705
},
{
"epoch": 0.2882205513784461,
"grad_norm": 0.9703027895486921,
"learning_rate": 1.9e-05,
"loss": 0.5571,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.59378981590271,
"step": 115,
"valid_targets_mean": 1433.0,
"valid_targets_min": 641
},
{
"epoch": 0.3007518796992481,
"grad_norm": 1.0870045224468745,
"learning_rate": 1.9833333333333335e-05,
"loss": 0.5618,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5384276509284973,
"step": 120,
"valid_targets_mean": 1152.8,
"valid_targets_min": 707
},
{
"epoch": 0.3132832080200501,
"grad_norm": 0.7700331714467691,
"learning_rate": 2.066666666666667e-05,
"loss": 0.5313,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4870373010635376,
"step": 125,
"valid_targets_mean": 1642.6,
"valid_targets_min": 668
},
{
"epoch": 0.3258145363408521,
"grad_norm": 0.8789807975640997,
"learning_rate": 2.15e-05,
"loss": 0.55,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5609132051467896,
"step": 130,
"valid_targets_mean": 1823.2,
"valid_targets_min": 723
},
{
"epoch": 0.3383458646616541,
"grad_norm": 0.8680952778590917,
"learning_rate": 2.2333333333333335e-05,
"loss": 0.5295,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5506467223167419,
"step": 135,
"valid_targets_mean": 1608.2,
"valid_targets_min": 695
},
{
"epoch": 0.3508771929824561,
"grad_norm": 0.8541140586653626,
"learning_rate": 2.316666666666667e-05,
"loss": 0.5193,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5101579427719116,
"step": 140,
"valid_targets_mean": 1443.7,
"valid_targets_min": 700
},
{
"epoch": 0.3634085213032581,
"grad_norm": 0.8260671930156039,
"learning_rate": 2.4e-05,
"loss": 0.5157,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4936548173427582,
"step": 145,
"valid_targets_mean": 1672.0,
"valid_targets_min": 652
},
{
"epoch": 0.37593984962406013,
"grad_norm": 0.8136337955135831,
"learning_rate": 2.4833333333333335e-05,
"loss": 0.4955,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.48357129096984863,
"step": 150,
"valid_targets_mean": 1661.6,
"valid_targets_min": 512
},
{
"epoch": 0.38847117794486213,
"grad_norm": 0.8534689528421746,
"learning_rate": 2.566666666666667e-05,
"loss": 0.5144,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.49319130182266235,
"step": 155,
"valid_targets_mean": 1579.5,
"valid_targets_min": 759
},
{
"epoch": 0.40100250626566414,
"grad_norm": 0.883526035160319,
"learning_rate": 2.65e-05,
"loss": 0.5072,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5153460502624512,
"step": 160,
"valid_targets_mean": 1388.1,
"valid_targets_min": 620
},
{
"epoch": 0.41353383458646614,
"grad_norm": 0.8934224689157233,
"learning_rate": 2.7333333333333335e-05,
"loss": 0.5193,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5074542760848999,
"step": 165,
"valid_targets_mean": 1612.6,
"valid_targets_min": 852
},
{
"epoch": 0.42606516290726815,
"grad_norm": 0.9616691914227905,
"learning_rate": 2.8166666666666673e-05,
"loss": 0.5136,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5238911509513855,
"step": 170,
"valid_targets_mean": 1431.7,
"valid_targets_min": 700
},
{
"epoch": 0.43859649122807015,
"grad_norm": 0.9858740124742356,
"learning_rate": 2.9e-05,
"loss": 0.4963,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5187824368476868,
"step": 175,
"valid_targets_mean": 1465.6,
"valid_targets_min": 741
},
{
"epoch": 0.45112781954887216,
"grad_norm": 0.735291369280417,
"learning_rate": 2.9833333333333338e-05,
"loss": 0.4942,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.45208507776260376,
"step": 180,
"valid_targets_mean": 2039.2,
"valid_targets_min": 791
},
{
"epoch": 0.46365914786967416,
"grad_norm": 1.0047344686962825,
"learning_rate": 3.066666666666667e-05,
"loss": 0.5027,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5117394924163818,
"step": 185,
"valid_targets_mean": 1217.4,
"valid_targets_min": 661
},
{
"epoch": 0.47619047619047616,
"grad_norm": 0.9220228098812638,
"learning_rate": 3.15e-05,
"loss": 0.5381,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5018247365951538,
"step": 190,
"valid_targets_mean": 1535.4,
"valid_targets_min": 719
},
{
"epoch": 0.48872180451127817,
"grad_norm": 0.876119425855413,
"learning_rate": 3.233333333333334e-05,
"loss": 0.5283,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.47728070616722107,
"step": 195,
"valid_targets_mean": 1649.7,
"valid_targets_min": 757
},
{
"epoch": 0.5012531328320802,
"grad_norm": 0.9997111593508858,
"learning_rate": 3.316666666666667e-05,
"loss": 0.5187,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5628472566604614,
"step": 200,
"valid_targets_mean": 1353.6,
"valid_targets_min": 763
},
{
"epoch": 0.5137844611528822,
"grad_norm": 0.8132225371914904,
"learning_rate": 3.4e-05,
"loss": 0.5012,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5150524377822876,
"step": 205,
"valid_targets_mean": 1791.1,
"valid_targets_min": 846
},
{
"epoch": 0.5263157894736842,
"grad_norm": 0.9327330282905936,
"learning_rate": 3.483333333333334e-05,
"loss": 0.4954,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.532800018787384,
"step": 210,
"valid_targets_mean": 1568.7,
"valid_targets_min": 618
},
{
"epoch": 0.5388471177944862,
"grad_norm": 0.8846848263288521,
"learning_rate": 3.566666666666667e-05,
"loss": 0.4663,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4891921877861023,
"step": 215,
"valid_targets_mean": 1425.6,
"valid_targets_min": 710
},
{
"epoch": 0.5513784461152882,
"grad_norm": 0.9695082520487255,
"learning_rate": 3.65e-05,
"loss": 0.4982,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5132033824920654,
"step": 220,
"valid_targets_mean": 1346.6,
"valid_targets_min": 864
},
{
"epoch": 0.5639097744360902,
"grad_norm": 0.8581508906819361,
"learning_rate": 3.733333333333334e-05,
"loss": 0.4923,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4817158877849579,
"step": 225,
"valid_targets_mean": 1754.4,
"valid_targets_min": 779
},
{
"epoch": 0.5764411027568922,
"grad_norm": 0.9214561305546174,
"learning_rate": 3.8166666666666675e-05,
"loss": 0.502,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.44954267144203186,
"step": 230,
"valid_targets_mean": 1653.5,
"valid_targets_min": 704
},
{
"epoch": 0.5889724310776943,
"grad_norm": 0.9242189734635586,
"learning_rate": 3.9e-05,
"loss": 0.5113,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5104361772537231,
"step": 235,
"valid_targets_mean": 1506.8,
"valid_targets_min": 637
},
{
"epoch": 0.6015037593984962,
"grad_norm": 0.8447888551101427,
"learning_rate": 3.983333333333334e-05,
"loss": 0.4779,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4449302554130554,
"step": 240,
"valid_targets_mean": 1498.4,
"valid_targets_min": 641
},
{
"epoch": 0.6140350877192983,
"grad_norm": 0.7911386578048909,
"learning_rate": 3.9999659648947195e-05,
"loss": 0.4974,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.46942049264907837,
"step": 245,
"valid_targets_mean": 1783.9,
"valid_targets_min": 696
},
{
"epoch": 0.6265664160401002,
"grad_norm": 0.8948760865225036,
"learning_rate": 3.999827699264838e-05,
"loss": 0.4816,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.45563480257987976,
"step": 250,
"valid_targets_mean": 1652.1,
"valid_targets_min": 619
},
{
"epoch": 0.6390977443609023,
"grad_norm": 1.0798792286240466,
"learning_rate": 3.999583083263554e-05,
"loss": 0.5077,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5357744693756104,
"step": 255,
"valid_targets_mean": 1125.9,
"valid_targets_min": 604
},
{
"epoch": 0.6516290726817042,
"grad_norm": 0.8575070696270298,
"learning_rate": 3.999232129899488e-05,
"loss": 0.485,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4918670952320099,
"step": 260,
"valid_targets_mean": 1625.6,
"valid_targets_min": 680
},
{
"epoch": 0.6641604010025063,
"grad_norm": 0.8850776630304149,
"learning_rate": 3.99877485783626e-05,
"loss": 0.5075,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4874800145626068,
"step": 265,
"valid_targets_mean": 1550.4,
"valid_targets_min": 706
},
{
"epoch": 0.6766917293233082,
"grad_norm": 0.8166417216262739,
"learning_rate": 3.998211291391491e-05,
"loss": 0.4939,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.466120183467865,
"step": 270,
"valid_targets_mean": 1398.1,
"valid_targets_min": 715
},
{
"epoch": 0.6892230576441103,
"grad_norm": 0.7154428431730908,
"learning_rate": 3.997541460535513e-05,
"loss": 0.4851,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.423753023147583,
"step": 275,
"valid_targets_mean": 2016.4,
"valid_targets_min": 750
},
{
"epoch": 0.7017543859649122,
"grad_norm": 0.9213430720357552,
"learning_rate": 3.996765400889775e-05,
"loss": 0.4881,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5173033475875854,
"step": 280,
"valid_targets_mean": 1301.8,
"valid_targets_min": 515
},
{
"epoch": 0.7142857142857143,
"grad_norm": 0.8023640529446733,
"learning_rate": 3.9958831537249484e-05,
"loss": 0.4717,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4825734496116638,
"step": 285,
"valid_targets_mean": 1722.1,
"valid_targets_min": 726
},
{
"epoch": 0.7268170426065163,
"grad_norm": 0.8706455117416396,
"learning_rate": 3.9948947659587336e-05,
"loss": 0.4857,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.48450374603271484,
"step": 290,
"valid_targets_mean": 1486.9,
"valid_targets_min": 766
},
{
"epoch": 0.7393483709273183,
"grad_norm": 0.853787916696495,
"learning_rate": 3.993800290153359e-05,
"loss": 0.4717,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.46074461936950684,
"step": 295,
"valid_targets_mean": 1472.6,
"valid_targets_min": 735
},
{
"epoch": 0.7518796992481203,
"grad_norm": 0.7835226381075628,
"learning_rate": 3.992599784512795e-05,
"loss": 0.4826,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4899185001850128,
"step": 300,
"valid_targets_mean": 1684.3,
"valid_targets_min": 734
},
{
"epoch": 0.7644110275689223,
"grad_norm": 0.9897427541429156,
"learning_rate": 3.991293312879652e-05,
"loss": 0.4896,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5004957914352417,
"step": 305,
"valid_targets_mean": 1242.6,
"valid_targets_min": 606
},
{
"epoch": 0.7769423558897243,
"grad_norm": 0.8938793841595182,
"learning_rate": 3.989880944731786e-05,
"loss": 0.4727,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.49720150232315063,
"step": 310,
"valid_targets_mean": 1363.1,
"valid_targets_min": 624
},
{
"epoch": 0.7894736842105263,
"grad_norm": 0.8382612026880438,
"learning_rate": 3.9883627551786074e-05,
"loss": 0.4978,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.48912370204925537,
"step": 315,
"valid_targets_mean": 1447.2,
"valid_targets_min": 783
},
{
"epoch": 0.8020050125313283,
"grad_norm": 0.761439572529583,
"learning_rate": 3.9867388249570836e-05,
"loss": 0.4772,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4344802498817444,
"step": 320,
"valid_targets_mean": 1695.3,
"valid_targets_min": 632
},
{
"epoch": 0.8145363408521303,
"grad_norm": 0.8785902546234519,
"learning_rate": 3.985009240427443e-05,
"loss": 0.4843,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4630719721317291,
"step": 325,
"valid_targets_mean": 1378.1,
"valid_targets_min": 683
},
{
"epoch": 0.8270676691729323,
"grad_norm": 0.8228485065620595,
"learning_rate": 3.983174093568591e-05,
"loss": 0.4878,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.47758162021636963,
"step": 330,
"valid_targets_mean": 1509.3,
"valid_targets_min": 689
},
{
"epoch": 0.8395989974937343,
"grad_norm": 0.9717758816642885,
"learning_rate": 3.98123348197321e-05,
"loss": 0.4698,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4755571484565735,
"step": 335,
"valid_targets_mean": 1212.8,
"valid_targets_min": 613
},
{
"epoch": 0.8521303258145363,
"grad_norm": 0.8614023270458415,
"learning_rate": 3.979187508842571e-05,
"loss": 0.4786,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5126999020576477,
"step": 340,
"valid_targets_mean": 1392.4,
"valid_targets_min": 682
},
{
"epoch": 0.8646616541353384,
"grad_norm": 0.7611158699746358,
"learning_rate": 3.977036282981051e-05,
"loss": 0.471,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4539802372455597,
"step": 345,
"valid_targets_mean": 1647.4,
"valid_targets_min": 580
},
{
"epoch": 0.8771929824561403,
"grad_norm": 0.8616664739071654,
"learning_rate": 3.974779918790338e-05,
"loss": 0.4884,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.47764351963996887,
"step": 350,
"valid_targets_mean": 1359.4,
"valid_targets_min": 692
},
{
"epoch": 0.8897243107769424,
"grad_norm": 0.8389852758904627,
"learning_rate": 3.972418536263355e-05,
"loss": 0.4866,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5188552141189575,
"step": 355,
"valid_targets_mean": 1837.6,
"valid_targets_min": 776
},
{
"epoch": 0.9022556390977443,
"grad_norm": 0.8061384574451791,
"learning_rate": 3.969952260977877e-05,
"loss": 0.4801,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.48911938071250916,
"step": 360,
"valid_targets_mean": 1631.5,
"valid_targets_min": 600
},
{
"epoch": 0.9147869674185464,
"grad_norm": 0.6985984050928506,
"learning_rate": 3.9673812240898466e-05,
"loss": 0.4723,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.44051438570022583,
"step": 365,
"valid_targets_mean": 1888.6,
"valid_targets_min": 661
},
{
"epoch": 0.9273182957393483,
"grad_norm": 0.8231367554229954,
"learning_rate": 3.964705562326408e-05,
"loss": 0.472,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4927082359790802,
"step": 370,
"valid_targets_mean": 1530.6,
"valid_targets_min": 648
},
{
"epoch": 0.9398496240601504,
"grad_norm": 0.7775064944719399,
"learning_rate": 3.961925417978632e-05,
"loss": 0.4734,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4817471206188202,
"step": 375,
"valid_targets_mean": 1528.3,
"valid_targets_min": 806
},
{
"epoch": 0.9523809523809523,
"grad_norm": 0.7068701184233374,
"learning_rate": 3.959040938893946e-05,
"loss": 0.4798,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4508395791053772,
"step": 380,
"valid_targets_mean": 1779.4,
"valid_targets_min": 722
},
{
"epoch": 0.9649122807017544,
"grad_norm": 0.7292299060821251,
"learning_rate": 3.9560522784682766e-05,
"loss": 0.4798,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.45591917634010315,
"step": 385,
"valid_targets_mean": 1567.5,
"valid_targets_min": 769
},
{
"epoch": 0.9774436090225563,
"grad_norm": 0.8273251382462157,
"learning_rate": 3.952959595637889e-05,
"loss": 0.4629,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.46951186656951904,
"step": 390,
"valid_targets_mean": 1494.0,
"valid_targets_min": 732
},
{
"epoch": 0.9899749373433584,
"grad_norm": 0.8808510863928986,
"learning_rate": 3.9497630548709375e-05,
"loss": 0.4756,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5034153461456299,
"step": 395,
"valid_targets_mean": 1323.8,
"valid_targets_min": 646
},
{
"epoch": 1.0025062656641603,
"grad_norm": 1.3955002841045798,
"learning_rate": 3.946462826158714e-05,
"loss": 0.4594,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.45724648237228394,
"step": 400,
"valid_targets_mean": 1202.4,
"valid_targets_min": 772
},
{
"epoch": 1.0150375939849625,
"grad_norm": 0.8269517494031545,
"learning_rate": 3.943059085006613e-05,
"loss": 0.4671,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.45007988810539246,
"step": 405,
"valid_targets_mean": 1828.4,
"valid_targets_min": 547
},
{
"epoch": 1.0275689223057645,
"grad_norm": 0.7493476492079604,
"learning_rate": 3.9395520124247984e-05,
"loss": 0.4473,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4048207104206085,
"step": 410,
"valid_targets_mean": 1577.7,
"valid_targets_min": 512
},
{
"epoch": 1.0401002506265664,
"grad_norm": 0.750270123972134,
"learning_rate": 3.935941794918572e-05,
"loss": 0.4573,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4433751702308655,
"step": 415,
"valid_targets_mean": 1736.9,
"valid_targets_min": 813
},
{
"epoch": 1.0526315789473684,
"grad_norm": 0.834001389405996,
"learning_rate": 3.9322286244784597e-05,
"loss": 0.4473,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4450203776359558,
"step": 420,
"valid_targets_mean": 1364.6,
"valid_targets_min": 617
},
{
"epoch": 1.0651629072681703,
"grad_norm": 0.794005813742166,
"learning_rate": 3.9284126985700016e-05,
"loss": 0.4351,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41918760538101196,
"step": 425,
"valid_targets_mean": 1459.9,
"valid_targets_min": 712
},
{
"epoch": 1.0776942355889725,
"grad_norm": 0.68082444015136,
"learning_rate": 3.9244942201232507e-05,
"loss": 0.4397,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.43921422958374023,
"step": 430,
"valid_targets_mean": 1819.3,
"valid_targets_min": 691
},
{
"epoch": 1.0902255639097744,
"grad_norm": 0.7208521260635171,
"learning_rate": 3.9204733975219754e-05,
"loss": 0.4409,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.43605563044548035,
"step": 435,
"valid_targets_mean": 1719.1,
"valid_targets_min": 770
},
{
"epoch": 1.1027568922305764,
"grad_norm": 0.7489888034624764,
"learning_rate": 3.9163504445925865e-05,
"loss": 0.44,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.42712393403053284,
"step": 440,
"valid_targets_mean": 1795.5,
"valid_targets_min": 740
},
{
"epoch": 1.1152882205513786,
"grad_norm": 0.8335448902554116,
"learning_rate": 3.9121255805927615e-05,
"loss": 0.4535,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4607224464416504,
"step": 445,
"valid_targets_mean": 1583.9,
"valid_targets_min": 582
},
{
"epoch": 1.1278195488721805,
"grad_norm": 0.7654806876894515,
"learning_rate": 3.907799030199784e-05,
"loss": 0.4491,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.45577600598335266,
"step": 450,
"valid_targets_mean": 1676.8,
"valid_targets_min": 497
},
{
"epoch": 1.1403508771929824,
"grad_norm": 0.929732543345461,
"learning_rate": 3.903371023498596e-05,
"loss": 0.4441,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4645495116710663,
"step": 455,
"valid_targets_mean": 1188.7,
"valid_targets_min": 427
},
{
"epoch": 1.1528822055137844,
"grad_norm": 0.7852422253164664,
"learning_rate": 3.898841795969563e-05,
"loss": 0.4581,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4440934658050537,
"step": 460,
"valid_targets_mean": 1563.7,
"valid_targets_min": 649
},
{
"epoch": 1.1654135338345863,
"grad_norm": 0.8757302453873744,
"learning_rate": 3.8942115884759505e-05,
"loss": 0.4516,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4700010418891907,
"step": 465,
"valid_targets_mean": 1330.1,
"valid_targets_min": 607
},
{
"epoch": 1.1779448621553885,
"grad_norm": 0.8499909272201124,
"learning_rate": 3.889480647251115e-05,
"loss": 0.4419,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.441257119178772,
"step": 470,
"valid_targets_mean": 1289.9,
"valid_targets_min": 711
},
{
"epoch": 1.1904761904761905,
"grad_norm": 0.8713204781666626,
"learning_rate": 3.884649223885409e-05,
"loss": 0.4443,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4358166456222534,
"step": 475,
"valid_targets_mean": 1215.9,
"valid_targets_min": 638
},
{
"epoch": 1.2030075187969924,
"grad_norm": 0.876755138815846,
"learning_rate": 3.879717575312802e-05,
"loss": 0.4547,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4684372842311859,
"step": 480,
"valid_targets_mean": 1500.8,
"valid_targets_min": 691
},
{
"epoch": 1.2155388471177946,
"grad_norm": 0.8095343697428509,
"learning_rate": 3.874685963797218e-05,
"loss": 0.4396,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4964834749698639,
"step": 485,
"valid_targets_mean": 1463.1,
"valid_targets_min": 616
},
{
"epoch": 1.2280701754385965,
"grad_norm": 0.7915449220424059,
"learning_rate": 3.869554656918584e-05,
"loss": 0.4585,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.46632125973701477,
"step": 490,
"valid_targets_mean": 1562.4,
"valid_targets_min": 618
},
{
"epoch": 1.2406015037593985,
"grad_norm": 0.8865984056691865,
"learning_rate": 3.864323927558606e-05,
"loss": 0.4422,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4493522047996521,
"step": 495,
"valid_targets_mean": 1433.1,
"valid_targets_min": 736
},
{
"epoch": 1.2531328320802004,
"grad_norm": 0.8529626057872378,
"learning_rate": 3.858994053886254e-05,
"loss": 0.4557,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4794151782989502,
"step": 500,
"valid_targets_mean": 1461.8,
"valid_targets_min": 747
},
{
"epoch": 1.2656641604010024,
"grad_norm": 0.73988664696522,
"learning_rate": 3.853565319342968e-05,
"loss": 0.4464,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4382275938987732,
"step": 505,
"valid_targets_mean": 1613.5,
"valid_targets_min": 740
},
{
"epoch": 1.2781954887218046,
"grad_norm": 0.8499742272557709,
"learning_rate": 3.848038012627587e-05,
"loss": 0.4403,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.45188936591148376,
"step": 510,
"valid_targets_mean": 1327.7,
"valid_targets_min": 646
},
{
"epoch": 1.2907268170426065,
"grad_norm": 0.837806554951509,
"learning_rate": 3.8424124276809956e-05,
"loss": 0.4637,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5095597505569458,
"step": 515,
"valid_targets_mean": 1469.3,
"valid_targets_min": 898
},
{
"epoch": 1.3032581453634084,
"grad_norm": 0.7019492892336352,
"learning_rate": 3.8366888636704916e-05,
"loss": 0.4481,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4414467215538025,
"step": 520,
"valid_targets_mean": 1896.4,
"valid_targets_min": 809
},
{
"epoch": 1.3157894736842106,
"grad_norm": 0.8718791272496377,
"learning_rate": 3.830867624973875e-05,
"loss": 0.4268,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.43715381622314453,
"step": 525,
"valid_targets_mean": 1746.9,
"valid_targets_min": 756
},
{
"epoch": 1.3283208020050126,
"grad_norm": 0.8909245916987222,
"learning_rate": 3.824949021163265e-05,
"loss": 0.4586,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4303131699562073,
"step": 530,
"valid_targets_mean": 1252.4,
"valid_targets_min": 735
},
{
"epoch": 1.3408521303258145,
"grad_norm": 0.740400513063558,
"learning_rate": 3.8189333669886354e-05,
"loss": 0.4349,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4084957242012024,
"step": 535,
"valid_targets_mean": 1536.8,
"valid_targets_min": 612
},
{
"epoch": 1.3533834586466165,
"grad_norm": 0.8115498137086447,
"learning_rate": 3.8128209823610735e-05,
"loss": 0.4547,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.45533132553100586,
"step": 540,
"valid_targets_mean": 1381.9,
"valid_targets_min": 675
},
{
"epoch": 1.3659147869674184,
"grad_norm": 0.6549907366136642,
"learning_rate": 3.806612192335769e-05,
"loss": 0.4426,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.434446781873703,
"step": 545,
"valid_targets_mean": 1967.2,
"valid_targets_min": 689
},
{
"epoch": 1.3784461152882206,
"grad_norm": 0.7386898782090013,
"learning_rate": 3.800307327094733e-05,
"loss": 0.433,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41058215498924255,
"step": 550,
"valid_targets_mean": 1503.0,
"valid_targets_min": 626
},
{
"epoch": 1.3909774436090225,
"grad_norm": 0.6970870873869073,
"learning_rate": 3.7939067219292284e-05,
"loss": 0.4199,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.42085281014442444,
"step": 555,
"valid_targets_mean": 1788.9,
"valid_targets_min": 641
},
{
"epoch": 1.4035087719298245,
"grad_norm": 0.7828068884431777,
"learning_rate": 3.787410717221948e-05,
"loss": 0.4458,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.44116562604904175,
"step": 560,
"valid_targets_mean": 1510.1,
"valid_targets_min": 591
},
{
"epoch": 1.4160401002506267,
"grad_norm": 0.657058991321496,
"learning_rate": 3.780819658428911e-05,
"loss": 0.4276,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3789742588996887,
"step": 565,
"valid_targets_mean": 1835.7,
"valid_targets_min": 890
},
{
"epoch": 1.4285714285714286,
"grad_norm": 0.6395861985446157,
"learning_rate": 3.7741338960610885e-05,
"loss": 0.426,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3952099680900574,
"step": 570,
"valid_targets_mean": 2071.3,
"valid_targets_min": 612
},
{
"epoch": 1.4411027568922306,
"grad_norm": 0.8347795153118129,
"learning_rate": 3.767353785665765e-05,
"loss": 0.4436,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4351791441440582,
"step": 575,
"valid_targets_mean": 1441.6,
"valid_targets_min": 678
},
{
"epoch": 1.4536340852130325,
"grad_norm": 0.8033860871708556,
"learning_rate": 3.760479687807635e-05,
"loss": 0.4476,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.46652624011039734,
"step": 580,
"valid_targets_mean": 1368.7,
"valid_targets_min": 643
},
{
"epoch": 1.4661654135338344,
"grad_norm": 0.6504013496758371,
"learning_rate": 3.753511968049622e-05,
"loss": 0.4294,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4036969542503357,
"step": 585,
"valid_targets_mean": 1799.1,
"valid_targets_min": 431
},
{
"epoch": 1.4786967418546366,
"grad_norm": 0.7381764886100423,
"learning_rate": 3.746450996933438e-05,
"loss": 0.4513,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4756660461425781,
"step": 590,
"valid_targets_mean": 1778.2,
"valid_targets_min": 696
},
{
"epoch": 1.4912280701754386,
"grad_norm": 0.6936827960015854,
"learning_rate": 3.739297149959884e-05,
"loss": 0.4447,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.42643094062805176,
"step": 595,
"valid_targets_mean": 1805.2,
"valid_targets_min": 776
},
{
"epoch": 1.5037593984962405,
"grad_norm": 0.697264472740107,
"learning_rate": 3.732050807568878e-05,
"loss": 0.4349,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.44037938117980957,
"step": 600,
"valid_targets_mean": 1873.8,
"valid_targets_min": 780
},
{
"epoch": 1.5162907268170427,
"grad_norm": 0.7975402798314732,
"learning_rate": 3.724712355119218e-05,
"loss": 0.4468,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.45365750789642334,
"step": 605,
"valid_targets_mean": 1470.1,
"valid_targets_min": 710
},
{
"epoch": 1.5288220551378446,
"grad_norm": 0.6786451129500303,
"learning_rate": 3.7172821828681e-05,
"loss": 0.4255,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41390031576156616,
"step": 610,
"valid_targets_mean": 1679.1,
"valid_targets_min": 684
},
{
"epoch": 1.5413533834586466,
"grad_norm": 0.7498724478327179,
"learning_rate": 3.7097606859503514e-05,
"loss": 0.4409,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4342663884162903,
"step": 615,
"valid_targets_mean": 1787.7,
"valid_targets_min": 549
},
{
"epoch": 1.5538847117794488,
"grad_norm": 0.7293841788979415,
"learning_rate": 3.702148264357428e-05,
"loss": 0.4501,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.43075722455978394,
"step": 620,
"valid_targets_mean": 1639.2,
"valid_targets_min": 662
},
{
"epoch": 1.5664160401002505,
"grad_norm": 0.7467516913324442,
"learning_rate": 3.694445322916138e-05,
"loss": 0.4488,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4477841556072235,
"step": 625,
"valid_targets_mean": 1638.9,
"valid_targets_min": 763
},
{
"epoch": 1.5789473684210527,
"grad_norm": 0.719672038793378,
"learning_rate": 3.686652271267114e-05,
"loss": 0.4334,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.432794988155365,
"step": 630,
"valid_targets_mean": 1743.4,
"valid_targets_min": 722
},
{
"epoch": 1.5914786967418546,
"grad_norm": 0.8220616262947018,
"learning_rate": 3.678769523843026e-05,
"loss": 0.4415,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.48474329710006714,
"step": 635,
"valid_targets_mean": 1533.6,
"valid_targets_min": 673
},
{
"epoch": 1.6040100250626566,
"grad_norm": 0.7771307016733878,
"learning_rate": 3.6707974998465495e-05,
"loss": 0.4383,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4561769962310791,
"step": 640,
"valid_targets_mean": 1554.4,
"valid_targets_min": 745
},
{
"epoch": 1.6165413533834587,
"grad_norm": 0.7761963492870784,
"learning_rate": 3.662736623228062e-05,
"loss": 0.4391,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4495946764945984,
"step": 645,
"valid_targets_mean": 1472.9,
"valid_targets_min": 682
},
{
"epoch": 1.6290726817042607,
"grad_norm": 0.9978606274160214,
"learning_rate": 3.654587322663106e-05,
"loss": 0.4431,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4660862684249878,
"step": 650,
"valid_targets_mean": 1242.4,
"valid_targets_min": 668
},
{
"epoch": 1.6416040100250626,
"grad_norm": 0.8231848952496575,
"learning_rate": 3.64635003152959e-05,
"loss": 0.4522,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.45088180899620056,
"step": 655,
"valid_targets_mean": 1370.2,
"valid_targets_min": 623
},
{
"epoch": 1.6541353383458648,
"grad_norm": 0.7471399307147799,
"learning_rate": 3.6380251878847356e-05,
"loss": 0.4316,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4242357611656189,
"step": 660,
"valid_targets_mean": 1712.2,
"valid_targets_min": 708
},
{
"epoch": 1.6666666666666665,
"grad_norm": 0.6863054821104536,
"learning_rate": 3.62961323444179e-05,
"loss": 0.4625,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4189673066139221,
"step": 665,
"valid_targets_mean": 1814.9,
"valid_targets_min": 751
},
{
"epoch": 1.6791979949874687,
"grad_norm": 0.786683608287003,
"learning_rate": 3.62111461854648e-05,
"loss": 0.4414,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41071224212646484,
"step": 670,
"valid_targets_mean": 1481.6,
"valid_targets_min": 792
},
{
"epoch": 1.6917293233082706,
"grad_norm": 0.7691848685754727,
"learning_rate": 3.6125297921532195e-05,
"loss": 0.4441,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4370245337486267,
"step": 675,
"valid_targets_mean": 1699.9,
"valid_targets_min": 725
},
{
"epoch": 1.7042606516290726,
"grad_norm": 0.9076573617201255,
"learning_rate": 3.603859211801076e-05,
"loss": 0.4384,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4542667269706726,
"step": 680,
"valid_targets_mean": 1348.2,
"valid_targets_min": 759
},
{
"epoch": 1.7167919799498748,
"grad_norm": 0.7736607294399541,
"learning_rate": 3.5951033385894955e-05,
"loss": 0.4534,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41055262088775635,
"step": 685,
"valid_targets_mean": 1575.7,
"valid_targets_min": 739
},
{
"epoch": 1.7293233082706767,
"grad_norm": 0.7320170764193196,
"learning_rate": 3.5862626381537753e-05,
"loss": 0.4496,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.45462262630462646,
"step": 690,
"valid_targets_mean": 1573.2,
"valid_targets_min": 759
},
{
"epoch": 1.7418546365914787,
"grad_norm": 0.6977308735657218,
"learning_rate": 3.577337580640307e-05,
"loss": 0.4405,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.43793463706970215,
"step": 695,
"valid_targets_mean": 1780.3,
"valid_targets_min": 710
},
{
"epoch": 1.7543859649122808,
"grad_norm": 0.7573922216037817,
"learning_rate": 3.568328640681571e-05,
"loss": 0.4495,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4407915771007538,
"step": 700,
"valid_targets_mean": 1433.2,
"valid_targets_min": 668
},
{
"epoch": 1.7669172932330826,
"grad_norm": 0.6368027104315123,
"learning_rate": 3.559236297370896e-05,
"loss": 0.4414,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.40819916129112244,
"step": 705,
"valid_targets_mean": 1925.6,
"valid_targets_min": 697
},
{
"epoch": 1.7794486215538847,
"grad_norm": 0.8715670762508443,
"learning_rate": 3.550061034236982e-05,
"loss": 0.4544,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4552498459815979,
"step": 710,
"valid_targets_mean": 1135.5,
"valid_targets_min": 764
},
{
"epoch": 1.7919799498746867,
"grad_norm": 0.6732436597197207,
"learning_rate": 3.540803339218187e-05,
"loss": 0.4169,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.42018571496009827,
"step": 715,
"valid_targets_mean": 1718.8,
"valid_targets_min": 833
},
{
"epoch": 1.8045112781954886,
"grad_norm": 0.6263088871441661,
"learning_rate": 3.531463704636577e-05,
"loss": 0.4383,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4200010895729065,
"step": 720,
"valid_targets_mean": 2012.9,
"valid_targets_min": 767
},
{
"epoch": 1.8170426065162908,
"grad_norm": 0.8384719503511425,
"learning_rate": 3.5220426271717426e-05,
"loss": 0.4509,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4631063938140869,
"step": 725,
"valid_targets_mean": 1250.4,
"valid_targets_min": 679
},
{
"epoch": 1.8295739348370927,
"grad_norm": 0.6843860284326129,
"learning_rate": 3.512540607834391e-05,
"loss": 0.4309,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4209910035133362,
"step": 730,
"valid_targets_mean": 1574.2,
"valid_targets_min": 487
},
{
"epoch": 1.8421052631578947,
"grad_norm": 0.7167917125977807,
"learning_rate": 3.5029581519396973e-05,
"loss": 0.4364,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.45028942823410034,
"step": 735,
"valid_targets_mean": 1659.6,
"valid_targets_min": 847
},
{
"epoch": 1.8546365914786969,
"grad_norm": 0.7314433156494289,
"learning_rate": 3.4932957690804356e-05,
"loss": 0.437,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4612088203430176,
"step": 740,
"valid_targets_mean": 1709.0,
"valid_targets_min": 657
},
{
"epoch": 1.8671679197994986,
"grad_norm": 0.6529714800972096,
"learning_rate": 3.483553973099876e-05,
"loss": 0.4396,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.42435121536254883,
"step": 745,
"valid_targets_mean": 1877.4,
"valid_targets_min": 907
},
{
"epoch": 1.8796992481203008,
"grad_norm": 0.8085177111679603,
"learning_rate": 3.473733282064461e-05,
"loss": 0.4494,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41868555545806885,
"step": 750,
"valid_targets_mean": 1269.4,
"valid_targets_min": 695
},
{
"epoch": 1.8922305764411027,
"grad_norm": 0.6608685301684564,
"learning_rate": 3.463834218236253e-05,
"loss": 0.4278,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41869646310806274,
"step": 755,
"valid_targets_mean": 1862.9,
"valid_targets_min": 556
},
{
"epoch": 1.9047619047619047,
"grad_norm": 0.7928065426796191,
"learning_rate": 3.453857308045161e-05,
"loss": 0.4357,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.45367372035980225,
"step": 760,
"valid_targets_mean": 1294.9,
"valid_targets_min": 637
},
{
"epoch": 1.9172932330827068,
"grad_norm": 0.6788143401778731,
"learning_rate": 3.443803082060946e-05,
"loss": 0.4262,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3976025879383087,
"step": 765,
"valid_targets_mean": 1658.4,
"valid_targets_min": 820
},
{
"epoch": 1.9298245614035088,
"grad_norm": 0.6538240412408628,
"learning_rate": 3.433672074965006e-05,
"loss": 0.4287,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39592060446739197,
"step": 770,
"valid_targets_mean": 1820.4,
"valid_targets_min": 779
},
{
"epoch": 1.9423558897243107,
"grad_norm": 0.8563678164989981,
"learning_rate": 3.423464825521937e-05,
"loss": 0.4369,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.45248812437057495,
"step": 775,
"valid_targets_mean": 1235.8,
"valid_targets_min": 635
},
{
"epoch": 1.954887218045113,
"grad_norm": 0.7053812027905907,
"learning_rate": 3.41318187655089e-05,
"loss": 0.4271,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41953131556510925,
"step": 780,
"valid_targets_mean": 1648.8,
"valid_targets_min": 641
},
{
"epoch": 1.9674185463659146,
"grad_norm": 0.6560343734177334,
"learning_rate": 3.4028237748966964e-05,
"loss": 0.4302,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4066466689109802,
"step": 785,
"valid_targets_mean": 1783.5,
"valid_targets_min": 723
},
{
"epoch": 1.9799498746867168,
"grad_norm": 0.71779268417739,
"learning_rate": 3.3923910714007896e-05,
"loss": 0.4229,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41889142990112305,
"step": 790,
"valid_targets_mean": 1806.1,
"valid_targets_min": 618
},
{
"epoch": 1.9924812030075187,
"grad_norm": 0.7925201428753242,
"learning_rate": 3.381884320871912e-05,
"loss": 0.446,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.44433343410491943,
"step": 795,
"valid_targets_mean": 1272.4,
"valid_targets_min": 660
},
{
"epoch": 2.0050125313283207,
"grad_norm": 0.6808799015569391,
"learning_rate": 3.3713040820566126e-05,
"loss": 0.4209,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3874009847640991,
"step": 800,
"valid_targets_mean": 1936.3,
"valid_targets_min": 614
},
{
"epoch": 2.017543859649123,
"grad_norm": 0.7283657760207235,
"learning_rate": 3.360650917609526e-05,
"loss": 0.4173,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4246293008327484,
"step": 805,
"valid_targets_mean": 1771.4,
"valid_targets_min": 675
},
{
"epoch": 2.030075187969925,
"grad_norm": 0.7242169851088842,
"learning_rate": 3.349925394063458e-05,
"loss": 0.3962,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3861202597618103,
"step": 810,
"valid_targets_mean": 1551.2,
"valid_targets_min": 759
},
{
"epoch": 2.0426065162907268,
"grad_norm": 0.7303228725795594,
"learning_rate": 3.339128081799253e-05,
"loss": 0.3989,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3861742913722992,
"step": 815,
"valid_targets_mean": 1691.4,
"valid_targets_min": 606
},
{
"epoch": 2.055137844611529,
"grad_norm": 0.6855353676041959,
"learning_rate": 3.3282595550154626e-05,
"loss": 0.4015,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.37186822295188904,
"step": 820,
"valid_targets_mean": 1639.1,
"valid_targets_min": 653
},
{
"epoch": 2.0676691729323307,
"grad_norm": 0.692410129531128,
"learning_rate": 3.317320391697811e-05,
"loss": 0.4083,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38654983043670654,
"step": 825,
"valid_targets_mean": 1639.6,
"valid_targets_min": 649
},
{
"epoch": 2.080200501253133,
"grad_norm": 0.7399489397821283,
"learning_rate": 3.306311173588457e-05,
"loss": 0.414,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3871074318885803,
"step": 830,
"valid_targets_mean": 1683.2,
"valid_targets_min": 732
},
{
"epoch": 2.092731829573935,
"grad_norm": 0.6913493221032149,
"learning_rate": 3.2952324861550555e-05,
"loss": 0.4165,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3757636249065399,
"step": 835,
"valid_targets_mean": 1750.5,
"valid_targets_min": 761
},
{
"epoch": 2.1052631578947367,
"grad_norm": 0.7411241018642346,
"learning_rate": 3.284084918559625e-05,
"loss": 0.3957,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3886348605155945,
"step": 840,
"valid_targets_mean": 1481.9,
"valid_targets_min": 626
},
{
"epoch": 2.117794486215539,
"grad_norm": 0.7262456607700576,
"learning_rate": 3.2728690636272146e-05,
"loss": 0.4199,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39896824955940247,
"step": 845,
"valid_targets_mean": 1553.8,
"valid_targets_min": 597
},
{
"epoch": 2.1303258145363406,
"grad_norm": 0.7934064754386024,
"learning_rate": 3.261585517814378e-05,
"loss": 0.4075,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.37877142429351807,
"step": 850,
"valid_targets_mean": 1613.4,
"valid_targets_min": 634
},
{
"epoch": 2.142857142857143,
"grad_norm": 0.8938513100589157,
"learning_rate": 3.250234881177453e-05,
"loss": 0.4098,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41370391845703125,
"step": 855,
"valid_targets_mean": 1570.8,
"valid_targets_min": 787
},
{
"epoch": 2.155388471177945,
"grad_norm": 0.8636183813830631,
"learning_rate": 3.238817757340655e-05,
"loss": 0.4124,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4297131896018982,
"step": 860,
"valid_targets_mean": 1242.0,
"valid_targets_min": 647
},
{
"epoch": 2.1679197994987467,
"grad_norm": 0.8040657962250063,
"learning_rate": 3.2273347534639705e-05,
"loss": 0.4093,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3994751572608948,
"step": 865,
"valid_targets_mean": 1337.7,
"valid_targets_min": 604
},
{
"epoch": 2.180451127819549,
"grad_norm": 0.8949819852190363,
"learning_rate": 3.215786480210872e-05,
"loss": 0.4109,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.44259727001190186,
"step": 870,
"valid_targets_mean": 1190.2,
"valid_targets_min": 612
},
{
"epoch": 2.192982456140351,
"grad_norm": 0.7843761916464653,
"learning_rate": 3.204173551715841e-05,
"loss": 0.4103,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.40686869621276855,
"step": 875,
"valid_targets_mean": 1634.0,
"valid_targets_min": 736
},
{
"epoch": 2.2055137844611528,
"grad_norm": 0.9181123085943382,
"learning_rate": 3.1924965855517135e-05,
"loss": 0.4168,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4497694969177246,
"step": 880,
"valid_targets_mean": 1156.1,
"valid_targets_min": 647
},
{
"epoch": 2.218045112781955,
"grad_norm": 0.9016745290827513,
"learning_rate": 3.180756202696829e-05,
"loss": 0.4237,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41279345750808716,
"step": 885,
"valid_targets_mean": 1226.4,
"valid_targets_min": 641
},
{
"epoch": 2.230576441102757,
"grad_norm": 0.8164305088369335,
"learning_rate": 3.1689530275020144e-05,
"loss": 0.423,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4257521331310272,
"step": 890,
"valid_targets_mean": 1375.3,
"valid_targets_min": 759
},
{
"epoch": 2.243107769423559,
"grad_norm": 0.7873273298467938,
"learning_rate": 3.1570876876573784e-05,
"loss": 0.4244,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4395800232887268,
"step": 895,
"valid_targets_mean": 1549.6,
"valid_targets_min": 751
},
{
"epoch": 2.255639097744361,
"grad_norm": 0.7340510236447725,
"learning_rate": 3.145160814158932e-05,
"loss": 0.406,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4071933925151825,
"step": 900,
"valid_targets_mean": 1613.9,
"valid_targets_min": 631
},
{
"epoch": 2.2681704260651627,
"grad_norm": 0.7520496021818598,
"learning_rate": 3.13317304127503e-05,
"loss": 0.4019,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3831161558628082,
"step": 905,
"valid_targets_mean": 1568.0,
"valid_targets_min": 744
},
{
"epoch": 2.280701754385965,
"grad_norm": 0.7830050957892405,
"learning_rate": 3.121125006512644e-05,
"loss": 0.4211,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.40112242102622986,
"step": 910,
"valid_targets_mean": 1376.7,
"valid_targets_min": 599
},
{
"epoch": 2.293233082706767,
"grad_norm": 0.8046622817084864,
"learning_rate": 3.109017350583455e-05,
"loss": 0.4095,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4121268093585968,
"step": 915,
"valid_targets_mean": 1299.8,
"valid_targets_min": 662
},
{
"epoch": 2.305764411027569,
"grad_norm": 0.7944019436211966,
"learning_rate": 3.0968507173697895e-05,
"loss": 0.398,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4244152009487152,
"step": 920,
"valid_targets_mean": 1441.0,
"valid_targets_min": 549
},
{
"epoch": 2.318295739348371,
"grad_norm": 0.7834254798584847,
"learning_rate": 3.0846257538903664e-05,
"loss": 0.4027,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41384357213974,
"step": 925,
"valid_targets_mean": 1396.5,
"valid_targets_min": 736
},
{
"epoch": 2.3308270676691727,
"grad_norm": 0.8554976494228926,
"learning_rate": 3.072343110265898e-05,
"loss": 0.4015,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4069422483444214,
"step": 930,
"valid_targets_mean": 1266.4,
"valid_targets_min": 751
},
{
"epoch": 2.343358395989975,
"grad_norm": 0.7346165996567373,
"learning_rate": 3.060003439684512e-05,
"loss": 0.3969,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41480642557144165,
"step": 935,
"valid_targets_mean": 1756.4,
"valid_targets_min": 780
},
{
"epoch": 2.355889724310777,
"grad_norm": 0.7421530852305365,
"learning_rate": 3.047607398367017e-05,
"loss": 0.4009,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.383364737033844,
"step": 940,
"valid_targets_mean": 1633.2,
"valid_targets_min": 658
},
{
"epoch": 2.3684210526315788,
"grad_norm": 0.6801690596484246,
"learning_rate": 3.0351556455320035e-05,
"loss": 0.4018,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4127517640590668,
"step": 945,
"valid_targets_mean": 1938.9,
"valid_targets_min": 582
},
{
"epoch": 2.380952380952381,
"grad_norm": 0.7725948108990202,
"learning_rate": 3.0226488433607895e-05,
"loss": 0.4051,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4012241065502167,
"step": 950,
"valid_targets_mean": 1393.2,
"valid_targets_min": 580
},
{
"epoch": 2.393483709273183,
"grad_norm": 0.6915866301278406,
"learning_rate": 3.0100876569622003e-05,
"loss": 0.4127,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3731994032859802,
"step": 955,
"valid_targets_mean": 1624.6,
"valid_targets_min": 616
},
{
"epoch": 2.406015037593985,
"grad_norm": 0.8576673007047129,
"learning_rate": 2.9974727543372046e-05,
"loss": 0.3954,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4130919575691223,
"step": 960,
"valid_targets_mean": 1161.0,
"valid_targets_min": 615
},
{
"epoch": 2.418546365914787,
"grad_norm": 0.8074845008674979,
"learning_rate": 2.984804806343387e-05,
"loss": 0.4344,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4470018446445465,
"step": 965,
"valid_targets_mean": 1505.1,
"valid_targets_min": 750
},
{
"epoch": 2.431077694235589,
"grad_norm": 0.8067633327978935,
"learning_rate": 2.9720844866592726e-05,
"loss": 0.4337,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41012510657310486,
"step": 970,
"valid_targets_mean": 1421.9,
"valid_targets_min": 712
},
{
"epoch": 2.443609022556391,
"grad_norm": 0.6741517814058764,
"learning_rate": 2.959312471748499e-05,
"loss": 0.4035,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4045283794403076,
"step": 975,
"valid_targets_mean": 1926.6,
"valid_targets_min": 792
},
{
"epoch": 2.456140350877193,
"grad_norm": 0.7515096216376789,
"learning_rate": 2.946489440823846e-05,
"loss": 0.4097,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4114806652069092,
"step": 980,
"valid_targets_mean": 1532.7,
"valid_targets_min": 627
},
{
"epoch": 2.468671679197995,
"grad_norm": 0.9200860671355993,
"learning_rate": 2.9336160758111126e-05,
"loss": 0.4104,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3898894488811493,
"step": 985,
"valid_targets_mean": 1654.8,
"valid_targets_min": 803
},
{
"epoch": 2.481203007518797,
"grad_norm": 0.7828468543626956,
"learning_rate": 2.9206930613128528e-05,
"loss": 0.3933,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4030555486679077,
"step": 990,
"valid_targets_mean": 1597.6,
"valid_targets_min": 614
},
{
"epoch": 2.493734335839599,
"grad_norm": 0.7537348447247004,
"learning_rate": 2.9077210845719684e-05,
"loss": 0.4043,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3929010331630707,
"step": 995,
"valid_targets_mean": 1445.5,
"valid_targets_min": 740
},
{
"epoch": 2.506265664160401,
"grad_norm": 0.8806951653745704,
"learning_rate": 2.894700835435162e-05,
"loss": 0.4061,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4137513041496277,
"step": 1000,
"valid_targets_mean": 1148.5,
"valid_targets_min": 594
},
{
"epoch": 2.518796992481203,
"grad_norm": 0.7768143066148521,
"learning_rate": 2.8816330063162508e-05,
"loss": 0.4052,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4147866666316986,
"step": 1005,
"valid_targets_mean": 1484.4,
"valid_targets_min": 569
},
{
"epoch": 2.5313283208020048,
"grad_norm": 0.7017072219607325,
"learning_rate": 2.8685182921593458e-05,
"loss": 0.4007,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3714948296546936,
"step": 1010,
"valid_targets_mean": 1581.1,
"valid_targets_min": 739
},
{
"epoch": 2.543859649122807,
"grad_norm": 0.843310856599826,
"learning_rate": 2.85535739040189e-05,
"loss": 0.3918,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4192495346069336,
"step": 1015,
"valid_targets_mean": 1274.7,
"valid_targets_min": 712
},
{
"epoch": 2.556390977443609,
"grad_norm": 0.6861456371527429,
"learning_rate": 2.8421510009375767e-05,
"loss": 0.4024,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3807116746902466,
"step": 1020,
"valid_targets_mean": 1922.1,
"valid_targets_min": 606
},
{
"epoch": 2.568922305764411,
"grad_norm": 0.7797110243365486,
"learning_rate": 2.8288998260791217e-05,
"loss": 0.411,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.42045170068740845,
"step": 1025,
"valid_targets_mean": 1480.0,
"valid_targets_min": 738
},
{
"epoch": 2.581453634085213,
"grad_norm": 0.7493284725461535,
"learning_rate": 2.8156045705209182e-05,
"loss": 0.4207,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4190444350242615,
"step": 1030,
"valid_targets_mean": 1863.0,
"valid_targets_min": 686
},
{
"epoch": 2.593984962406015,
"grad_norm": 0.9715778306128012,
"learning_rate": 2.8022659413015616e-05,
"loss": 0.4098,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3925316035747528,
"step": 1035,
"valid_targets_mean": 1929.9,
"valid_targets_min": 803
},
{
"epoch": 2.606516290726817,
"grad_norm": 0.8202300357445593,
"learning_rate": 2.7888846477662474e-05,
"loss": 0.4054,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.435418039560318,
"step": 1040,
"valid_targets_mean": 1275.8,
"valid_targets_min": 627
},
{
"epoch": 2.619047619047619,
"grad_norm": 0.8065052554671248,
"learning_rate": 2.7754614015290512e-05,
"loss": 0.4052,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39296644926071167,
"step": 1045,
"valid_targets_mean": 1280.7,
"valid_targets_min": 625
},
{
"epoch": 2.6315789473684212,
"grad_norm": 0.8191455525882082,
"learning_rate": 2.7619969164350814e-05,
"loss": 0.4182,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.42794761061668396,
"step": 1050,
"valid_targets_mean": 1448.4,
"valid_targets_min": 633
},
{
"epoch": 2.644110275689223,
"grad_norm": 0.7770712433544891,
"learning_rate": 2.7484919085225203e-05,
"loss": 0.4021,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4223942756652832,
"step": 1055,
"valid_targets_mean": 1644.5,
"valid_targets_min": 625
},
{
"epoch": 2.656641604010025,
"grad_norm": 0.6581892363783577,
"learning_rate": 2.734947095984544e-05,
"loss": 0.4125,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36657726764678955,
"step": 1060,
"valid_targets_mean": 1759.4,
"valid_targets_min": 739
},
{
"epoch": 2.6691729323308273,
"grad_norm": 0.8646402657014103,
"learning_rate": 2.7213631991311303e-05,
"loss": 0.4111,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4207683801651001,
"step": 1065,
"valid_targets_mean": 1353.9,
"valid_targets_min": 734
},
{
"epoch": 2.681704260651629,
"grad_norm": 0.7319122083607305,
"learning_rate": 2.7077409403507503e-05,
"loss": 0.414,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.43628567457199097,
"step": 1070,
"valid_targets_mean": 1707.5,
"valid_targets_min": 697
},
{
"epoch": 2.694235588972431,
"grad_norm": 0.8342526774577462,
"learning_rate": 2.6940810440719545e-05,
"loss": 0.4015,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41509440541267395,
"step": 1075,
"valid_targets_mean": 1300.9,
"valid_targets_min": 700
},
{
"epoch": 2.706766917293233,
"grad_norm": 0.7902962051936262,
"learning_rate": 2.680384236724846e-05,
"loss": 0.4186,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.40989580750465393,
"step": 1080,
"valid_targets_mean": 1366.7,
"valid_targets_min": 773
},
{
"epoch": 2.719298245614035,
"grad_norm": 0.8189188800470997,
"learning_rate": 2.6666512467024506e-05,
"loss": 0.4085,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.40571969747543335,
"step": 1085,
"valid_targets_mean": 1528.2,
"valid_targets_min": 678
},
{
"epoch": 2.731829573934837,
"grad_norm": 0.6678143387997657,
"learning_rate": 2.6528828043219798e-05,
"loss": 0.3835,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35410448908805847,
"step": 1090,
"valid_targets_mean": 1824.2,
"valid_targets_min": 755
},
{
"epoch": 2.744360902255639,
"grad_norm": 0.8132896008113664,
"learning_rate": 2.6390796417859937e-05,
"loss": 0.4163,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4377764165401459,
"step": 1095,
"valid_targets_mean": 1452.2,
"valid_targets_min": 593
},
{
"epoch": 2.756892230576441,
"grad_norm": 0.8586131814893146,
"learning_rate": 2.625242493143462e-05,
"loss": 0.3996,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.42037975788116455,
"step": 1100,
"valid_targets_mean": 1271.2,
"valid_targets_min": 735
},
{
"epoch": 2.769423558897243,
"grad_norm": 0.8058156340524054,
"learning_rate": 2.6113720942507277e-05,
"loss": 0.4146,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38332855701446533,
"step": 1105,
"valid_targets_mean": 1270.2,
"valid_targets_min": 771
},
{
"epoch": 2.781954887218045,
"grad_norm": 0.8219535969366415,
"learning_rate": 2.5974691827323747e-05,
"loss": 0.429,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.44158488512039185,
"step": 1110,
"valid_targets_mean": 1437.2,
"valid_targets_min": 730
},
{
"epoch": 2.7944862155388472,
"grad_norm": 0.7213100908994415,
"learning_rate": 2.5835344979419992e-05,
"loss": 0.3945,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4036768078804016,
"step": 1115,
"valid_targets_mean": 1581.8,
"valid_targets_min": 723
},
{
"epoch": 2.807017543859649,
"grad_norm": 0.7160583991847018,
"learning_rate": 2.569568780922895e-05,
"loss": 0.3934,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3890629708766937,
"step": 1120,
"valid_targets_mean": 1857.3,
"valid_targets_min": 736
},
{
"epoch": 2.819548872180451,
"grad_norm": 0.6030656035509802,
"learning_rate": 2.5555727743686404e-05,
"loss": 0.4076,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38233476877212524,
"step": 1125,
"valid_targets_mean": 2011.9,
"valid_targets_min": 805
},
{
"epoch": 2.8320802005012533,
"grad_norm": 0.7989881611675778,
"learning_rate": 2.541547222583606e-05,
"loss": 0.4064,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4262428283691406,
"step": 1130,
"valid_targets_mean": 1484.9,
"valid_targets_min": 637
},
{
"epoch": 2.844611528822055,
"grad_norm": 0.7670693348326154,
"learning_rate": 2.5274928714433694e-05,
"loss": 0.3995,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39012211561203003,
"step": 1135,
"valid_targets_mean": 1433.4,
"valid_targets_min": 727
},
{
"epoch": 2.857142857142857,
"grad_norm": 0.7329434894141582,
"learning_rate": 2.5134104683550514e-05,
"loss": 0.4124,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39894163608551025,
"step": 1140,
"valid_targets_mean": 1380.9,
"valid_targets_min": 740
},
{
"epoch": 2.8696741854636594,
"grad_norm": 0.8338871518868882,
"learning_rate": 2.4993007622175704e-05,
"loss": 0.4149,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4379524290561676,
"step": 1145,
"valid_targets_mean": 1150.1,
"valid_targets_min": 742
},
{
"epoch": 2.882205513784461,
"grad_norm": 0.8375405196939691,
"learning_rate": 2.485164503381813e-05,
"loss": 0.4183,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41192495822906494,
"step": 1150,
"valid_targets_mean": 1206.6,
"valid_targets_min": 741
},
{
"epoch": 2.8947368421052633,
"grad_norm": 0.8512040546281613,
"learning_rate": 2.471002443610732e-05,
"loss": 0.4053,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4261419475078583,
"step": 1155,
"valid_targets_mean": 1371.6,
"valid_targets_min": 753
},
{
"epoch": 2.907268170426065,
"grad_norm": 0.6905558161225763,
"learning_rate": 2.4568153360393692e-05,
"loss": 0.4029,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.399014413356781,
"step": 1160,
"valid_targets_mean": 1654.8,
"valid_targets_min": 690
},
{
"epoch": 2.919799498746867,
"grad_norm": 0.7795729327578536,
"learning_rate": 2.4426039351348016e-05,
"loss": 0.4074,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.43625450134277344,
"step": 1165,
"valid_targets_mean": 1567.2,
"valid_targets_min": 847
},
{
"epoch": 2.932330827067669,
"grad_norm": 0.696371052775749,
"learning_rate": 2.4283689966560205e-05,
"loss": 0.3931,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.37107014656066895,
"step": 1170,
"valid_targets_mean": 1794.1,
"valid_targets_min": 497
},
{
"epoch": 2.944862155388471,
"grad_norm": 0.7090326982911932,
"learning_rate": 2.414111277613739e-05,
"loss": 0.4125,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4178668260574341,
"step": 1175,
"valid_targets_mean": 1603.3,
"valid_targets_min": 580
},
{
"epoch": 2.9573934837092732,
"grad_norm": 0.7411555880347811,
"learning_rate": 2.3998315362301367e-05,
"loss": 0.4061,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39853695034980774,
"step": 1180,
"valid_targets_mean": 1641.4,
"valid_targets_min": 728
},
{
"epoch": 2.969924812030075,
"grad_norm": 0.7157497676079337,
"learning_rate": 2.3855305318985353e-05,
"loss": 0.3983,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36533060669898987,
"step": 1185,
"valid_targets_mean": 1622.5,
"valid_targets_min": 698
},
{
"epoch": 2.982456140350877,
"grad_norm": 0.8725579962324914,
"learning_rate": 2.371209025143016e-05,
"loss": 0.4121,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.43822264671325684,
"step": 1190,
"valid_targets_mean": 1269.9,
"valid_targets_min": 735
},
{
"epoch": 2.9949874686716793,
"grad_norm": 0.7833600243534485,
"learning_rate": 2.3568677775779716e-05,
"loss": 0.4112,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.42297977209091187,
"step": 1195,
"valid_targets_mean": 1463.1,
"valid_targets_min": 726
},
{
"epoch": 3.007518796992481,
"grad_norm": 0.7524818799186773,
"learning_rate": 2.34250755186761e-05,
"loss": 0.3949,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38546401262283325,
"step": 1200,
"valid_targets_mean": 1481.1,
"valid_targets_min": 427
},
{
"epoch": 3.020050125313283,
"grad_norm": 0.7269674684497573,
"learning_rate": 2.3281291116853903e-05,
"loss": 0.3829,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3603457808494568,
"step": 1205,
"valid_targets_mean": 1522.2,
"valid_targets_min": 739
},
{
"epoch": 3.0325814536340854,
"grad_norm": 0.8551480190784867,
"learning_rate": 2.3137332216734132e-05,
"loss": 0.3841,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39454448223114014,
"step": 1210,
"valid_targets_mean": 1317.9,
"valid_targets_min": 770
},
{
"epoch": 3.045112781954887,
"grad_norm": 0.7469871028238138,
"learning_rate": 2.2993206474017576e-05,
"loss": 0.3769,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.37015241384506226,
"step": 1215,
"valid_targets_mean": 1690.9,
"valid_targets_min": 641
},
{
"epoch": 3.0576441102756893,
"grad_norm": 0.7826410951147598,
"learning_rate": 2.2848921553277674e-05,
"loss": 0.3927,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3815900385379791,
"step": 1220,
"valid_targets_mean": 1497.1,
"valid_targets_min": 559
},
{
"epoch": 3.0701754385964914,
"grad_norm": 0.7933176189991209,
"learning_rate": 2.270448512755292e-05,
"loss": 0.3832,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38196152448654175,
"step": 1225,
"valid_targets_mean": 1607.9,
"valid_targets_min": 736
},
{
"epoch": 3.082706766917293,
"grad_norm": 0.790107574228091,
"learning_rate": 2.2559904877938803e-05,
"loss": 0.3789,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3487051725387573,
"step": 1230,
"valid_targets_mean": 1450.9,
"valid_targets_min": 790
},
{
"epoch": 3.0952380952380953,
"grad_norm": 0.8142328651019091,
"learning_rate": 2.241518849317933e-05,
"loss": 0.3952,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39621999859809875,
"step": 1235,
"valid_targets_mean": 1305.6,
"valid_targets_min": 644
},
{
"epoch": 3.107769423558897,
"grad_norm": 0.8379046959783162,
"learning_rate": 2.227034366925814e-05,
"loss": 0.3802,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4048861265182495,
"step": 1240,
"valid_targets_mean": 1384.4,
"valid_targets_min": 671
},
{
"epoch": 3.1203007518796992,
"grad_norm": 0.6842829184061702,
"learning_rate": 2.2125378108989257e-05,
"loss": 0.3814,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3356945514678955,
"step": 1245,
"valid_targets_mean": 1627.7,
"valid_targets_min": 674
},
{
"epoch": 3.1328320802005014,
"grad_norm": 0.8150081084171071,
"learning_rate": 2.198029952160742e-05,
"loss": 0.3809,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.37989962100982666,
"step": 1250,
"valid_targets_mean": 1446.6,
"valid_targets_min": 730
},
{
"epoch": 3.145363408521303,
"grad_norm": 0.9065745245822664,
"learning_rate": 2.1835115622358113e-05,
"loss": 0.378,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4172065258026123,
"step": 1255,
"valid_targets_mean": 1195.8,
"valid_targets_min": 753
},
{
"epoch": 3.1578947368421053,
"grad_norm": 0.797667764442672,
"learning_rate": 2.16898341320873e-05,
"loss": 0.3919,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.40342891216278076,
"step": 1260,
"valid_targets_mean": 1506.2,
"valid_targets_min": 731
},
{
"epoch": 3.170426065162907,
"grad_norm": 0.82920979257728,
"learning_rate": 2.154446277683081e-05,
"loss": 0.3809,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3811074495315552,
"step": 1265,
"valid_targets_mean": 1842.1,
"valid_targets_min": 785
},
{
"epoch": 3.182957393483709,
"grad_norm": 0.8609643118461479,
"learning_rate": 2.1399009287403464e-05,
"loss": 0.3845,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39563995599746704,
"step": 1270,
"valid_targets_mean": 1280.9,
"valid_targets_min": 648
},
{
"epoch": 3.1954887218045114,
"grad_norm": 0.8575392033562702,
"learning_rate": 2.1253481398987965e-05,
"loss": 0.4002,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3950216770172119,
"step": 1275,
"valid_targets_mean": 1358.1,
"valid_targets_min": 647
},
{
"epoch": 3.208020050125313,
"grad_norm": 0.9108739991923912,
"learning_rate": 2.1107886850723527e-05,
"loss": 0.3846,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4000109136104584,
"step": 1280,
"valid_targets_mean": 1106.9,
"valid_targets_min": 641
},
{
"epoch": 3.2205513784461153,
"grad_norm": 0.7468400623645985,
"learning_rate": 2.096223338529433e-05,
"loss": 0.3853,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3904936909675598,
"step": 1285,
"valid_targets_mean": 1848.0,
"valid_targets_min": 591
},
{
"epoch": 3.2330827067669174,
"grad_norm": 0.8186011370453243,
"learning_rate": 2.0816528748517753e-05,
"loss": 0.361,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36554673314094543,
"step": 1290,
"valid_targets_mean": 1278.4,
"valid_targets_min": 668
},
{
"epoch": 3.245614035087719,
"grad_norm": 0.8151988405085241,
"learning_rate": 2.067078068893245e-05,
"loss": 0.4028,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39774322509765625,
"step": 1295,
"valid_targets_mean": 1529.9,
"valid_targets_min": 658
},
{
"epoch": 3.2581453634085213,
"grad_norm": 0.8281756678211144,
"learning_rate": 2.0524996957386297e-05,
"loss": 0.3752,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3898346722126007,
"step": 1300,
"valid_targets_mean": 1593.6,
"valid_targets_min": 805
},
{
"epoch": 3.2706766917293235,
"grad_norm": 0.9193624245811687,
"learning_rate": 2.037918530662419e-05,
"loss": 0.3747,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3889794945716858,
"step": 1305,
"valid_targets_mean": 1275.1,
"valid_targets_min": 727
},
{
"epoch": 3.2832080200501252,
"grad_norm": 0.7060451326734182,
"learning_rate": 2.0233353490875773e-05,
"loss": 0.3758,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36215609312057495,
"step": 1310,
"valid_targets_mean": 1662.6,
"valid_targets_min": 906
},
{
"epoch": 3.2957393483709274,
"grad_norm": 0.855976457831653,
"learning_rate": 2.0087509265443038e-05,
"loss": 0.3621,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3756676912307739,
"step": 1315,
"valid_targets_mean": 1367.2,
"valid_targets_min": 610
},
{
"epoch": 3.308270676691729,
"grad_norm": 0.6704601020026827,
"learning_rate": 1.9941660386287946e-05,
"loss": 0.3722,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35007140040397644,
"step": 1320,
"valid_targets_mean": 1839.5,
"valid_targets_min": 863
},
{
"epoch": 3.3208020050125313,
"grad_norm": 0.8563097392456597,
"learning_rate": 1.9795814609619925e-05,
"loss": 0.3717,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3476110100746155,
"step": 1325,
"valid_targets_mean": 1556.3,
"valid_targets_min": 638
},
{
"epoch": 3.3333333333333335,
"grad_norm": 0.7150924024477262,
"learning_rate": 1.9649979691483426e-05,
"loss": 0.391,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35187599062919617,
"step": 1330,
"valid_targets_mean": 1596.5,
"valid_targets_min": 684
},
{
"epoch": 3.345864661654135,
"grad_norm": 0.874681446014307,
"learning_rate": 1.950416338734543e-05,
"loss": 0.3823,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38704127073287964,
"step": 1335,
"valid_targets_mean": 1349.2,
"valid_targets_min": 634
},
{
"epoch": 3.3583959899749374,
"grad_norm": 0.7127005137441507,
"learning_rate": 1.9358373451683047e-05,
"loss": 0.3748,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3557561933994293,
"step": 1340,
"valid_targets_mean": 1700.8,
"valid_targets_min": 772
},
{
"epoch": 3.370927318295739,
"grad_norm": 0.7447684168527092,
"learning_rate": 1.9212617637571104e-05,
"loss": 0.3857,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3649003505706787,
"step": 1345,
"valid_targets_mean": 1484.9,
"valid_targets_min": 720
},
{
"epoch": 3.3834586466165413,
"grad_norm": 0.7347851437358622,
"learning_rate": 1.906690369626986e-05,
"loss": 0.376,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3531128764152527,
"step": 1350,
"valid_targets_mean": 2112.2,
"valid_targets_min": 760
},
{
"epoch": 3.3959899749373434,
"grad_norm": 0.8064613290760294,
"learning_rate": 1.8921239376812782e-05,
"loss": 0.3851,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39800935983657837,
"step": 1355,
"valid_targets_mean": 1409.1,
"valid_targets_min": 547
},
{
"epoch": 3.408521303258145,
"grad_norm": 0.8320563236978611,
"learning_rate": 1.8775632425594467e-05,
"loss": 0.3764,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3904891610145569,
"step": 1360,
"valid_targets_mean": 1439.0,
"valid_targets_min": 582
},
{
"epoch": 3.4210526315789473,
"grad_norm": 0.7726671817431195,
"learning_rate": 1.8630090585958676e-05,
"loss": 0.3811,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35375556349754333,
"step": 1365,
"valid_targets_mean": 1437.6,
"valid_targets_min": 603
},
{
"epoch": 3.4335839598997495,
"grad_norm": 0.8558402291328209,
"learning_rate": 1.8484621597786547e-05,
"loss": 0.3817,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38245689868927,
"step": 1370,
"valid_targets_mean": 1224.4,
"valid_targets_min": 649
},
{
"epoch": 3.4461152882205512,
"grad_norm": 0.5757371027540783,
"learning_rate": 1.8339233197085006e-05,
"loss": 0.3628,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33190372586250305,
"step": 1375,
"valid_targets_mean": 2415.6,
"valid_targets_min": 710
},
{
"epoch": 3.4586466165413534,
"grad_norm": 0.825713016108861,
"learning_rate": 1.8193933115575347e-05,
"loss": 0.378,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3584468364715576,
"step": 1380,
"valid_targets_mean": 1350.2,
"valid_targets_min": 633
},
{
"epoch": 3.4711779448621556,
"grad_norm": 0.8287859805915134,
"learning_rate": 1.8048729080282076e-05,
"loss": 0.3875,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3999900221824646,
"step": 1385,
"valid_targets_mean": 1409.1,
"valid_targets_min": 614
},
{
"epoch": 3.4837092731829573,
"grad_norm": 0.8158738188623406,
"learning_rate": 1.7903628813121976e-05,
"loss": 0.3814,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3847970962524414,
"step": 1390,
"valid_targets_mean": 1375.8,
"valid_targets_min": 637
},
{
"epoch": 3.4962406015037595,
"grad_norm": 0.9806141114108778,
"learning_rate": 1.775864003049347e-05,
"loss": 0.3935,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.43756890296936035,
"step": 1395,
"valid_targets_mean": 1060.5,
"valid_targets_min": 662
},
{
"epoch": 3.5087719298245617,
"grad_norm": 0.9388820887180466,
"learning_rate": 1.761377044286626e-05,
"loss": 0.3796,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4040834307670593,
"step": 1400,
"valid_targets_mean": 1213.0,
"valid_targets_min": 647
},
{
"epoch": 3.5213032581453634,
"grad_norm": 0.7535144988237941,
"learning_rate": 1.746902775437129e-05,
"loss": 0.3795,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3977915048599243,
"step": 1405,
"valid_targets_mean": 1685.8,
"valid_targets_min": 750
},
{
"epoch": 3.5338345864661656,
"grad_norm": 0.8469541672098552,
"learning_rate": 1.7324419662391022e-05,
"loss": 0.3809,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41281846165657043,
"step": 1410,
"valid_targets_mean": 1293.9,
"valid_targets_min": 808
},
{
"epoch": 3.5463659147869673,
"grad_norm": 0.8324398561287292,
"learning_rate": 1.717995385715012e-05,
"loss": 0.3903,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4070686101913452,
"step": 1415,
"valid_targets_mean": 1355.4,
"valid_targets_min": 678
},
{
"epoch": 3.5588972431077694,
"grad_norm": 0.7587541092559444,
"learning_rate": 1.7035638021306467e-05,
"loss": 0.3842,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3555733561515808,
"step": 1420,
"valid_targets_mean": 1363.9,
"valid_targets_min": 732
},
{
"epoch": 3.571428571428571,
"grad_norm": 0.7000504384554054,
"learning_rate": 1.6891479829542613e-05,
"loss": 0.3612,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3316074013710022,
"step": 1425,
"valid_targets_mean": 1686.2,
"valid_targets_min": 618
},
{
"epoch": 3.5839598997493733,
"grad_norm": 0.75752037224634,
"learning_rate": 1.674748694815763e-05,
"loss": 0.3795,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3835802674293518,
"step": 1430,
"valid_targets_mean": 1537.1,
"valid_targets_min": 642
},
{
"epoch": 3.5964912280701755,
"grad_norm": 0.8010015007994762,
"learning_rate": 1.6603667034659427e-05,
"loss": 0.3928,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.411435067653656,
"step": 1435,
"valid_targets_mean": 1514.9,
"valid_targets_min": 728
},
{
"epoch": 3.6090225563909772,
"grad_norm": 0.8427953151697828,
"learning_rate": 1.6460027737357507e-05,
"loss": 0.3716,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.40870070457458496,
"step": 1440,
"valid_targets_mean": 1463.4,
"valid_targets_min": 606
},
{
"epoch": 3.6215538847117794,
"grad_norm": 0.7787782369841209,
"learning_rate": 1.631657669495626e-05,
"loss": 0.3733,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39166802167892456,
"step": 1445,
"valid_targets_mean": 1467.4,
"valid_targets_min": 612
},
{
"epoch": 3.6340852130325816,
"grad_norm": 0.6852579078522438,
"learning_rate": 1.617332153614872e-05,
"loss": 0.3887,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3678228557109833,
"step": 1450,
"valid_targets_mean": 1836.8,
"valid_targets_min": 696
},
{
"epoch": 3.6466165413533833,
"grad_norm": 0.7780223246726713,
"learning_rate": 1.6030269879210882e-05,
"loss": 0.386,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.40032103657722473,
"step": 1455,
"valid_targets_mean": 1652.6,
"valid_targets_min": 659
},
{
"epoch": 3.6591478696741855,
"grad_norm": 0.774087518297068,
"learning_rate": 1.588742933159654e-05,
"loss": 0.383,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.40035444498062134,
"step": 1460,
"valid_targets_mean": 1652.8,
"valid_targets_min": 812
},
{
"epoch": 3.6716791979949877,
"grad_norm": 0.8202131392244767,
"learning_rate": 1.5744807489532766e-05,
"loss": 0.3878,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3781554400920868,
"step": 1465,
"valid_targets_mean": 1362.3,
"valid_targets_min": 731
},
{
"epoch": 3.6842105263157894,
"grad_norm": 0.6274517417710915,
"learning_rate": 1.5602411937615913e-05,
"loss": 0.3786,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3454907238483429,
"step": 1470,
"valid_targets_mean": 2221.0,
"valid_targets_min": 735
},
{
"epoch": 3.6967418546365916,
"grad_norm": 0.7412975036909865,
"learning_rate": 1.5460250248408273e-05,
"loss": 0.3668,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3802959620952606,
"step": 1475,
"valid_targets_mean": 1712.1,
"valid_targets_min": 612
},
{
"epoch": 3.7092731829573937,
"grad_norm": 0.7855111216631802,
"learning_rate": 1.5318329982035377e-05,
"loss": 0.3818,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3868759870529175,
"step": 1480,
"valid_targets_mean": 1714.4,
"valid_targets_min": 700
},
{
"epoch": 3.7218045112781954,
"grad_norm": 0.736710536856686,
"learning_rate": 1.5176658685783947e-05,
"loss": 0.3855,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3668166399002075,
"step": 1485,
"valid_targets_mean": 1660.8,
"valid_targets_min": 446
},
{
"epoch": 3.7343358395989976,
"grad_norm": 0.8206766311128828,
"learning_rate": 1.5035243893700535e-05,
"loss": 0.3894,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.37208032608032227,
"step": 1490,
"valid_targets_mean": 1317.5,
"valid_targets_min": 803
},
{
"epoch": 3.7468671679197993,
"grad_norm": 0.7536710064360856,
"learning_rate": 1.489409312619085e-05,
"loss": 0.3802,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4111523926258087,
"step": 1495,
"valid_targets_mean": 1782.4,
"valid_targets_min": 743
},
{
"epoch": 3.7593984962406015,
"grad_norm": 0.8364316599116858,
"learning_rate": 1.4753213889619841e-05,
"loss": 0.3964,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.37741121649742126,
"step": 1500,
"valid_targets_mean": 1405.8,
"valid_targets_min": 689
},
{
"epoch": 3.7719298245614032,
"grad_norm": 0.7382628078568813,
"learning_rate": 1.4612613675912512e-05,
"loss": 0.3883,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3787643015384674,
"step": 1505,
"valid_targets_mean": 1712.3,
"valid_targets_min": 823
},
{
"epoch": 3.7844611528822054,
"grad_norm": 0.6910932736068953,
"learning_rate": 1.4472299962155492e-05,
"loss": 0.3769,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3468532860279083,
"step": 1510,
"valid_targets_mean": 1946.0,
"valid_targets_min": 828
},
{
"epoch": 3.7969924812030076,
"grad_norm": 0.8339264830378494,
"learning_rate": 1.4332280210199398e-05,
"loss": 0.3867,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3885452151298523,
"step": 1515,
"valid_targets_mean": 1297.6,
"valid_targets_min": 668
},
{
"epoch": 3.8095238095238093,
"grad_norm": 0.872717499618267,
"learning_rate": 1.4192561866262044e-05,
"loss": 0.3799,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4075925946235657,
"step": 1520,
"valid_targets_mean": 1278.0,
"valid_targets_min": 673
},
{
"epoch": 3.8220551378446115,
"grad_norm": 0.7816477296593115,
"learning_rate": 1.4053152360532427e-05,
"loss": 0.3719,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36789825558662415,
"step": 1525,
"valid_targets_mean": 1411.9,
"valid_targets_min": 512
},
{
"epoch": 3.8345864661654137,
"grad_norm": 0.8184377473400201,
"learning_rate": 1.3914059106775604e-05,
"loss": 0.3742,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39734745025634766,
"step": 1530,
"valid_targets_mean": 1389.1,
"valid_targets_min": 672
},
{
"epoch": 3.8471177944862154,
"grad_norm": 0.7515969959020206,
"learning_rate": 1.3775289501938416e-05,
"loss": 0.3778,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35744476318359375,
"step": 1535,
"valid_targets_mean": 1552.1,
"valid_targets_min": 735
},
{
"epoch": 3.8596491228070176,
"grad_norm": 1.402932466951827,
"learning_rate": 1.3636850925756142e-05,
"loss": 0.3929,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.398654580116272,
"step": 1540,
"valid_targets_mean": 1618.7,
"valid_targets_min": 763
},
{
"epoch": 3.8721804511278197,
"grad_norm": 0.7241112856924586,
"learning_rate": 1.3498750740360027e-05,
"loss": 0.3917,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3559287190437317,
"step": 1545,
"valid_targets_mean": 1582.4,
"valid_targets_min": 760
},
{
"epoch": 3.8847117794486214,
"grad_norm": 0.786140387459547,
"learning_rate": 1.3360996289885776e-05,
"loss": 0.3838,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4194572865962982,
"step": 1550,
"valid_targets_mean": 1645.4,
"valid_targets_min": 606
},
{
"epoch": 3.8972431077694236,
"grad_norm": 0.746696006161874,
"learning_rate": 1.322359490008299e-05,
"loss": 0.3742,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35523080825805664,
"step": 1555,
"valid_targets_mean": 1703.7,
"valid_targets_min": 646
},
{
"epoch": 3.909774436090226,
"grad_norm": 0.9846950145958907,
"learning_rate": 1.3086553877925585e-05,
"loss": 0.3714,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38349106907844543,
"step": 1560,
"valid_targets_mean": 1237.5,
"valid_targets_min": 705
},
{
"epoch": 3.9223057644110275,
"grad_norm": 0.8985016618941177,
"learning_rate": 1.2949880511223211e-05,
"loss": 0.3789,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4175274968147278,
"step": 1565,
"valid_targets_mean": 1333.2,
"valid_targets_min": 630
},
{
"epoch": 3.9348370927318297,
"grad_norm": 0.6911200210688109,
"learning_rate": 1.2813582068233692e-05,
"loss": 0.3831,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3444245159626007,
"step": 1570,
"valid_targets_mean": 1834.4,
"valid_targets_min": 697
},
{
"epoch": 3.9473684210526314,
"grad_norm": 0.7919585555911091,
"learning_rate": 1.267766579727648e-05,
"loss": 0.3713,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.389361172914505,
"step": 1575,
"valid_targets_mean": 1583.0,
"valid_targets_min": 836
},
{
"epoch": 3.9598997493734336,
"grad_norm": 0.8157469102876466,
"learning_rate": 1.2542138926347219e-05,
"loss": 0.379,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3919796049594879,
"step": 1580,
"valid_targets_mean": 1358.5,
"valid_targets_min": 680
},
{
"epoch": 3.9724310776942353,
"grad_norm": 0.7821339820816429,
"learning_rate": 1.2407008662733354e-05,
"loss": 0.3702,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3739144206047058,
"step": 1585,
"valid_targets_mean": 1608.3,
"valid_targets_min": 637
},
{
"epoch": 3.9849624060150375,
"grad_norm": 0.868931464158729,
"learning_rate": 1.227228219263084e-05,
"loss": 0.3906,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38923490047454834,
"step": 1590,
"valid_targets_mean": 1265.2,
"valid_targets_min": 569
},
{
"epoch": 3.9974937343358397,
"grad_norm": 0.8489429245422577,
"learning_rate": 1.213796668076198e-05,
"loss": 0.3873,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3736661970615387,
"step": 1595,
"valid_targets_mean": 1199.9,
"valid_targets_min": 675
},
{
"epoch": 4.010025062656641,
"grad_norm": 0.8281878570880064,
"learning_rate": 1.2004069269994428e-05,
"loss": 0.3628,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3639451563358307,
"step": 1600,
"valid_targets_mean": 1318.1,
"valid_targets_min": 770
},
{
"epoch": 4.022556390977444,
"grad_norm": 0.7895190631839228,
"learning_rate": 1.1870597080961308e-05,
"loss": 0.3557,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3259652853012085,
"step": 1605,
"valid_targets_mean": 1762.9,
"valid_targets_min": 642
},
{
"epoch": 4.035087719298246,
"grad_norm": 0.802426459799358,
"learning_rate": 1.173755721168256e-05,
"loss": 0.3618,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3714388310909271,
"step": 1610,
"valid_targets_mean": 1549.3,
"valid_targets_min": 690
},
{
"epoch": 4.0476190476190474,
"grad_norm": 0.8143506173545653,
"learning_rate": 1.160495673718744e-05,
"loss": 0.3525,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35231781005859375,
"step": 1615,
"valid_targets_mean": 1594.6,
"valid_targets_min": 712
},
{
"epoch": 4.06015037593985,
"grad_norm": 0.9111334577543276,
"learning_rate": 1.1472802709138335e-05,
"loss": 0.3706,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3951305150985718,
"step": 1620,
"valid_targets_mean": 1291.2,
"valid_targets_min": 668
},
{
"epoch": 4.072681704260652,
"grad_norm": 0.8425868781426432,
"learning_rate": 1.1341102155455656e-05,
"loss": 0.3578,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39576777815818787,
"step": 1625,
"valid_targets_mean": 1467.6,
"valid_targets_min": 749
},
{
"epoch": 4.0852130325814535,
"grad_norm": 0.7205858154186265,
"learning_rate": 1.1209862079944198e-05,
"loss": 0.3638,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34877854585647583,
"step": 1630,
"valid_targets_mean": 1918.1,
"valid_targets_min": 615
},
{
"epoch": 4.097744360902255,
"grad_norm": 0.817307445046329,
"learning_rate": 1.107908946192061e-05,
"loss": 0.3663,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3607422411441803,
"step": 1635,
"valid_targets_mean": 1482.1,
"valid_targets_min": 670
},
{
"epoch": 4.110275689223058,
"grad_norm": 0.6775906170894619,
"learning_rate": 1.094879125584228e-05,
"loss": 0.3534,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3350442051887512,
"step": 1640,
"valid_targets_mean": 2074.7,
"valid_targets_min": 768
},
{
"epoch": 4.12280701754386,
"grad_norm": 0.803939930552659,
"learning_rate": 1.081897439093746e-05,
"loss": 0.3625,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35282477736473083,
"step": 1645,
"valid_targets_mean": 1837.1,
"valid_targets_min": 623
},
{
"epoch": 4.135338345864661,
"grad_norm": 0.6719027490110735,
"learning_rate": 1.06896457708368e-05,
"loss": 0.3599,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30827414989471436,
"step": 1650,
"valid_targets_mean": 1859.6,
"valid_targets_min": 770
},
{
"epoch": 4.147869674185464,
"grad_norm": 0.8890942790175431,
"learning_rate": 1.0560812273206215e-05,
"loss": 0.3603,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35805994272232056,
"step": 1655,
"valid_targets_mean": 1179.4,
"valid_targets_min": 584
},
{
"epoch": 4.160401002506266,
"grad_norm": 0.9347570311165075,
"learning_rate": 1.0432480749381093e-05,
"loss": 0.3651,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38746821880340576,
"step": 1660,
"valid_targets_mean": 1310.9,
"valid_targets_min": 547
},
{
"epoch": 4.172932330827067,
"grad_norm": 0.7661173205906099,
"learning_rate": 1.0304658024002002e-05,
"loss": 0.3572,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36425966024398804,
"step": 1665,
"valid_targets_mean": 1738.4,
"valid_targets_min": 829
},
{
"epoch": 4.18546365914787,
"grad_norm": 0.8060874639241118,
"learning_rate": 1.0177350894651708e-05,
"loss": 0.3593,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35910189151763916,
"step": 1670,
"valid_targets_mean": 1575.8,
"valid_targets_min": 601
},
{
"epoch": 4.197994987468672,
"grad_norm": 0.8596879937640446,
"learning_rate": 1.005056613149371e-05,
"loss": 0.3636,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3838065564632416,
"step": 1675,
"valid_targets_mean": 1479.6,
"valid_targets_min": 809
},
{
"epoch": 4.2105263157894735,
"grad_norm": 0.8946285345544369,
"learning_rate": 9.924310476912196e-06,
"loss": 0.3631,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.37626439332962036,
"step": 1680,
"valid_targets_mean": 1487.1,
"valid_targets_min": 497
},
{
"epoch": 4.223057644110276,
"grad_norm": 0.8825349942645182,
"learning_rate": 9.798590645153473e-06,
"loss": 0.381,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3923253118991852,
"step": 1685,
"valid_targets_mean": 1455.2,
"valid_targets_min": 845
},
{
"epoch": 4.235588972431078,
"grad_norm": 0.93306267674495,
"learning_rate": 9.673413321968934e-06,
"loss": 0.3722,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38771435618400574,
"step": 1690,
"valid_targets_mean": 1282.9,
"valid_targets_min": 669
},
{
"epoch": 4.2481203007518795,
"grad_norm": 0.7779130043424681,
"learning_rate": 9.548785164259461e-06,
"loss": 0.3712,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3406516909599304,
"step": 1695,
"valid_targets_mean": 1632.7,
"valid_targets_min": 732
},
{
"epoch": 4.260651629072681,
"grad_norm": 0.8720036417507798,
"learning_rate": 9.424712799721472e-06,
"loss": 0.35,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33976292610168457,
"step": 1700,
"valid_targets_mean": 1479.0,
"valid_targets_min": 870
},
{
"epoch": 4.273182957393484,
"grad_norm": 0.7464452343967485,
"learning_rate": 9.301202826494428e-06,
"loss": 0.337,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3533982038497925,
"step": 1705,
"valid_targets_mean": 1797.8,
"valid_targets_min": 731
},
{
"epoch": 4.285714285714286,
"grad_norm": 0.7435681280302439,
"learning_rate": 9.178261812809934e-06,
"loss": 0.3678,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3589174747467041,
"step": 1710,
"valid_targets_mean": 1821.1,
"valid_targets_min": 726
},
{
"epoch": 4.298245614035087,
"grad_norm": 0.778889350577403,
"learning_rate": 9.055896296642481e-06,
"loss": 0.3518,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33703145384788513,
"step": 1715,
"valid_targets_mean": 1610.9,
"valid_targets_min": 675
},
{
"epoch": 4.31077694235589,
"grad_norm": 0.8506286418125527,
"learning_rate": 8.934112785361726e-06,
"loss": 0.3607,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34579887986183167,
"step": 1720,
"valid_targets_mean": 1332.0,
"valid_targets_min": 688
},
{
"epoch": 4.323308270676692,
"grad_norm": 0.9291089907595611,
"learning_rate": 8.81291775538644e-06,
"loss": 0.3556,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38011759519577026,
"step": 1725,
"valid_targets_mean": 1284.8,
"valid_targets_min": 630
},
{
"epoch": 4.335839598997493,
"grad_norm": 0.6690325258072783,
"learning_rate": 8.692317651840101e-06,
"loss": 0.3555,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32536908984184265,
"step": 1730,
"valid_targets_mean": 1997.7,
"valid_targets_min": 786
},
{
"epoch": 4.348370927318296,
"grad_norm": 0.9074516079828592,
"learning_rate": 8.572318888208135e-06,
"loss": 0.3676,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3661980628967285,
"step": 1735,
"valid_targets_mean": 1338.8,
"valid_targets_min": 616
},
{
"epoch": 4.360902255639098,
"grad_norm": 0.8527642895752466,
"learning_rate": 8.452927845996854e-06,
"loss": 0.362,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.37586355209350586,
"step": 1740,
"valid_targets_mean": 1442.7,
"valid_targets_min": 720
},
{
"epoch": 4.3734335839598995,
"grad_norm": 0.8219078557591193,
"learning_rate": 8.334150874394075e-06,
"loss": 0.3684,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3594920039176941,
"step": 1745,
"valid_targets_mean": 1493.9,
"valid_targets_min": 801
},
{
"epoch": 4.385964912280702,
"grad_norm": 0.8353204864275154,
"learning_rate": 8.215994289931495e-06,
"loss": 0.3576,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35578057169914246,
"step": 1750,
"valid_targets_mean": 1527.2,
"valid_targets_min": 541
},
{
"epoch": 4.398496240601504,
"grad_norm": 0.7956210259824289,
"learning_rate": 8.098464376148765e-06,
"loss": 0.3654,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3480678200721741,
"step": 1755,
"valid_targets_mean": 1577.4,
"valid_targets_min": 695
},
{
"epoch": 4.4110275689223055,
"grad_norm": 0.7360349691169649,
"learning_rate": 7.981567383259332e-06,
"loss": 0.3556,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3734129071235657,
"step": 1760,
"valid_targets_mean": 1928.4,
"valid_targets_min": 955
},
{
"epoch": 4.423558897243108,
"grad_norm": 0.7466972215988599,
"learning_rate": 7.865309527818063e-06,
"loss": 0.3588,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36618131399154663,
"step": 1765,
"valid_targets_mean": 1647.6,
"valid_targets_min": 687
},
{
"epoch": 4.43609022556391,
"grad_norm": 0.7337253381376839,
"learning_rate": 7.749696992390639e-06,
"loss": 0.3747,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35264647006988525,
"step": 1770,
"valid_targets_mean": 1900.6,
"valid_targets_min": 762
},
{
"epoch": 4.448621553884712,
"grad_norm": 0.8256948835805943,
"learning_rate": 7.634735925224775e-06,
"loss": 0.3592,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3599795401096344,
"step": 1775,
"valid_targets_mean": 1649.9,
"valid_targets_min": 759
},
{
"epoch": 4.461152882205514,
"grad_norm": 0.7054242837878253,
"learning_rate": 7.52043243992326e-06,
"loss": 0.359,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3563918173313141,
"step": 1780,
"valid_targets_mean": 1888.7,
"valid_targets_min": 648
},
{
"epoch": 4.473684210526316,
"grad_norm": 0.7727784292442016,
"learning_rate": 7.406792615118817e-06,
"loss": 0.3604,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34456440806388855,
"step": 1785,
"valid_targets_mean": 1713.8,
"valid_targets_min": 859
},
{
"epoch": 4.486215538847118,
"grad_norm": 0.899362312646131,
"learning_rate": 7.293822494150873e-06,
"loss": 0.3626,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3481305241584778,
"step": 1790,
"valid_targets_mean": 1350.6,
"valid_targets_min": 647
},
{
"epoch": 4.498746867167919,
"grad_norm": 1.0030617001629305,
"learning_rate": 7.181528084744158e-06,
"loss": 0.3756,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3706076145172119,
"step": 1795,
"valid_targets_mean": 1212.2,
"valid_targets_min": 813
},
{
"epoch": 4.511278195488722,
"grad_norm": 0.8685403109225178,
"learning_rate": 7.069915358689217e-06,
"loss": 0.3562,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35512638092041016,
"step": 1800,
"valid_targets_mean": 1506.2,
"valid_targets_min": 674
},
{
"epoch": 4.523809523809524,
"grad_norm": 0.791391902227632,
"learning_rate": 6.958990251524835e-06,
"loss": 0.3619,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3524198830127716,
"step": 1805,
"valid_targets_mean": 1708.7,
"valid_targets_min": 694
},
{
"epoch": 4.5363408521303255,
"grad_norm": 0.850539866019582,
"learning_rate": 6.848758662222383e-06,
"loss": 0.3581,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3824390172958374,
"step": 1810,
"valid_targets_mean": 1577.4,
"valid_targets_min": 696
},
{
"epoch": 4.548872180451128,
"grad_norm": 0.7417616786020198,
"learning_rate": 6.7392264528721115e-06,
"loss": 0.3572,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35435038805007935,
"step": 1815,
"valid_targets_mean": 1835.0,
"valid_targets_min": 641
},
{
"epoch": 4.56140350877193,
"grad_norm": 0.7298033849882459,
"learning_rate": 6.630399448371416e-06,
"loss": 0.3618,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3553701937198639,
"step": 1820,
"valid_targets_mean": 1834.9,
"valid_targets_min": 762
},
{
"epoch": 4.5739348370927315,
"grad_norm": 0.7386646813073992,
"learning_rate": 6.522283436115046e-06,
"loss": 0.3656,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34987321496009827,
"step": 1825,
"valid_targets_mean": 1897.1,
"valid_targets_min": 681
},
{
"epoch": 4.586466165413534,
"grad_norm": 0.9117333312529112,
"learning_rate": 6.4148841656873675e-06,
"loss": 0.365,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.37728363275527954,
"step": 1830,
"valid_targets_mean": 1299.4,
"valid_targets_min": 752
},
{
"epoch": 4.598997493734336,
"grad_norm": 0.8585767700709204,
"learning_rate": 6.30820734855657e-06,
"loss": 0.3608,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3458079695701599,
"step": 1835,
"valid_targets_mean": 1372.4,
"valid_targets_min": 607
},
{
"epoch": 4.611528822055138,
"grad_norm": 0.8891984018071964,
"learning_rate": 6.20225865777095e-06,
"loss": 0.3552,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3820083439350128,
"step": 1840,
"valid_targets_mean": 1449.1,
"valid_targets_min": 680
},
{
"epoch": 4.62406015037594,
"grad_norm": 0.8609457925615936,
"learning_rate": 6.097043727657217e-06,
"loss": 0.342,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3439421057701111,
"step": 1845,
"valid_targets_mean": 1339.6,
"valid_targets_min": 673
},
{
"epoch": 4.636591478696742,
"grad_norm": 0.7270110211202997,
"learning_rate": 5.992568153520857e-06,
"loss": 0.348,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.37063953280448914,
"step": 1850,
"valid_targets_mean": 2134.8,
"valid_targets_min": 765
},
{
"epoch": 4.649122807017544,
"grad_norm": 1.0500038015228268,
"learning_rate": 5.888837491348571e-06,
"loss": 0.371,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3767479360103607,
"step": 1855,
"valid_targets_mean": 1030.8,
"valid_targets_min": 573
},
{
"epoch": 4.661654135338345,
"grad_norm": 0.847254023798705,
"learning_rate": 5.785857257512832e-06,
"loss": 0.3629,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3626944422721863,
"step": 1860,
"valid_targets_mean": 1471.9,
"valid_targets_min": 527
},
{
"epoch": 4.674185463659148,
"grad_norm": 0.9635967545009322,
"learning_rate": 5.683632928478482e-06,
"loss": 0.3618,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3814685344696045,
"step": 1865,
"valid_targets_mean": 1185.8,
"valid_targets_min": 594
},
{
"epoch": 4.68671679197995,
"grad_norm": 0.9070771054657724,
"learning_rate": 5.582169940511544e-06,
"loss": 0.3541,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.40833860635757446,
"step": 1870,
"valid_targets_mean": 1570.8,
"valid_targets_min": 779
},
{
"epoch": 4.6992481203007515,
"grad_norm": 0.9110570136084019,
"learning_rate": 5.481473689390091e-06,
"loss": 0.3673,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3828704059123993,
"step": 1875,
"valid_targets_mean": 1317.1,
"valid_targets_min": 593
},
{
"epoch": 4.711779448621554,
"grad_norm": 0.7552289878852424,
"learning_rate": 5.381549530117316e-06,
"loss": 0.3644,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34786558151245117,
"step": 1880,
"valid_targets_mean": 1746.8,
"valid_targets_min": 684
},
{
"epoch": 4.724310776942356,
"grad_norm": 0.7974717999699457,
"learning_rate": 5.28240277663674e-06,
"loss": 0.3671,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3812224268913269,
"step": 1885,
"valid_targets_mean": 1763.7,
"valid_targets_min": 749
},
{
"epoch": 4.7368421052631575,
"grad_norm": 0.8018221996566985,
"learning_rate": 5.184038701549625e-06,
"loss": 0.3722,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34907105565071106,
"step": 1890,
"valid_targets_mean": 1535.9,
"valid_targets_min": 611
},
{
"epoch": 4.74937343358396,
"grad_norm": 1.0612817335584024,
"learning_rate": 5.0864625358345905e-06,
"loss": 0.3479,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3929625451564789,
"step": 1895,
"valid_targets_mean": 1216.8,
"valid_targets_min": 630
},
{
"epoch": 4.761904761904762,
"grad_norm": 0.8051135149931052,
"learning_rate": 4.989679468569397e-06,
"loss": 0.3629,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.363945871591568,
"step": 1900,
"valid_targets_mean": 1474.8,
"valid_targets_min": 620
},
{
"epoch": 4.774436090225564,
"grad_norm": 0.9182485899219046,
"learning_rate": 4.8936946466550314e-06,
"loss": 0.3677,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3759042024612427,
"step": 1905,
"valid_targets_mean": 1297.1,
"valid_targets_min": 743
},
{
"epoch": 4.786967418546366,
"grad_norm": 0.7657705069544768,
"learning_rate": 4.7985131745419745e-06,
"loss": 0.357,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32689371705055237,
"step": 1910,
"valid_targets_mean": 1497.1,
"valid_targets_min": 839
},
{
"epoch": 4.799498746867168,
"grad_norm": 0.8497598450408903,
"learning_rate": 4.70414011395875e-06,
"loss": 0.3765,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.37474486231803894,
"step": 1915,
"valid_targets_mean": 1450.2,
"valid_targets_min": 877
},
{
"epoch": 4.81203007518797,
"grad_norm": 0.7728175275071781,
"learning_rate": 4.610580483642748e-06,
"loss": 0.3415,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3417680263519287,
"step": 1920,
"valid_targets_mean": 1643.1,
"valid_targets_min": 546
},
{
"epoch": 4.824561403508772,
"grad_norm": 0.9532676991958982,
"learning_rate": 4.517839259073322e-06,
"loss": 0.3664,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36030858755111694,
"step": 1925,
"valid_targets_mean": 1275.1,
"valid_targets_min": 808
},
{
"epoch": 4.837092731829574,
"grad_norm": 0.9345157903304709,
"learning_rate": 4.425921372207201e-06,
"loss": 0.3461,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3792099952697754,
"step": 1930,
"valid_targets_mean": 1430.7,
"valid_targets_min": 624
},
{
"epoch": 4.849624060150376,
"grad_norm": 0.7539611030052837,
"learning_rate": 4.334831711216209e-06,
"loss": 0.3657,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34493887424468994,
"step": 1935,
"valid_targets_mean": 1758.8,
"valid_targets_min": 710
},
{
"epoch": 4.862155388471178,
"grad_norm": 0.9159733465774322,
"learning_rate": 4.244575120227303e-06,
"loss": 0.359,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36277949810028076,
"step": 1940,
"valid_targets_mean": 1285.8,
"valid_targets_min": 644
},
{
"epoch": 4.87468671679198,
"grad_norm": 0.9917940035130076,
"learning_rate": 4.155156399064977e-06,
"loss": 0.3698,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38035470247268677,
"step": 1945,
"valid_targets_mean": 1096.4,
"valid_targets_min": 714
},
{
"epoch": 4.887218045112782,
"grad_norm": 0.8015299461290246,
"learning_rate": 4.066580302996004e-06,
"loss": 0.3595,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3660500943660736,
"step": 1950,
"valid_targets_mean": 1779.2,
"valid_targets_min": 633
},
{
"epoch": 4.899749373433584,
"grad_norm": 0.7730826390487343,
"learning_rate": 3.978851542476547e-06,
"loss": 0.3477,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35395899415016174,
"step": 1955,
"valid_targets_mean": 1528.4,
"valid_targets_min": 717
},
{
"epoch": 4.912280701754386,
"grad_norm": 0.8322096620940707,
"learning_rate": 3.891974782901666e-06,
"loss": 0.3748,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3833672106266022,
"step": 1960,
"valid_targets_mean": 1523.0,
"valid_targets_min": 649
},
{
"epoch": 4.924812030075188,
"grad_norm": 0.8417191866894002,
"learning_rate": 3.805954644357206e-06,
"loss": 0.3562,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36396360397338867,
"step": 1965,
"valid_targets_mean": 1521.4,
"valid_targets_min": 677
},
{
"epoch": 4.93734335839599,
"grad_norm": 0.7291255713954147,
"learning_rate": 3.720795701374109e-06,
"loss": 0.3693,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3539201617240906,
"step": 1970,
"valid_targets_mean": 1917.8,
"valid_targets_min": 843
},
{
"epoch": 4.949874686716792,
"grad_norm": 0.765912521584318,
"learning_rate": 3.636502482685125e-06,
"loss": 0.359,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34452271461486816,
"step": 1975,
"valid_targets_mean": 1644.4,
"valid_targets_min": 790
},
{
"epoch": 4.962406015037594,
"grad_norm": 0.9446530016204546,
"learning_rate": 3.553079470984002e-06,
"loss": 0.3639,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36807239055633545,
"step": 1980,
"valid_targets_mean": 1425.4,
"valid_targets_min": 639
},
{
"epoch": 4.974937343358396,
"grad_norm": 0.865026454724945,
"learning_rate": 3.4705311026870848e-06,
"loss": 0.3639,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3669584393501282,
"step": 1985,
"valid_targets_mean": 1342.0,
"valid_targets_min": 616
},
{
"epoch": 4.987468671679198,
"grad_norm": 0.685032837279059,
"learning_rate": 3.3888617676973645e-06,
"loss": 0.3487,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3004816174507141,
"step": 1990,
"valid_targets_mean": 1637.9,
"valid_targets_min": 668
},
{
"epoch": 5.0,
"grad_norm": 0.7752928593373746,
"learning_rate": 3.3080758091710676e-06,
"loss": 0.3592,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32106631994247437,
"step": 1995,
"valid_targets_mean": 1756.9,
"valid_targets_min": 707
},
{
"epoch": 5.012531328320802,
"grad_norm": 0.7859655644778784,
"learning_rate": 3.22817752328666e-06,
"loss": 0.3555,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3647751808166504,
"step": 2000,
"valid_targets_mean": 1628.6,
"valid_targets_min": 698
},
{
"epoch": 5.025062656641604,
"grad_norm": 0.8381726272050811,
"learning_rate": 3.1491711590163777e-06,
"loss": 0.3522,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3480392098426819,
"step": 2005,
"valid_targets_mean": 1342.1,
"valid_targets_min": 710
},
{
"epoch": 5.037593984962406,
"grad_norm": 0.7949118559237107,
"learning_rate": 3.071060917900277e-06,
"loss": 0.3535,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36498141288757324,
"step": 2010,
"valid_targets_mean": 1858.8,
"valid_targets_min": 750
},
{
"epoch": 5.050125313283208,
"grad_norm": 0.7779825433802904,
"learning_rate": 2.9938509538227944e-06,
"loss": 0.3511,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31160634756088257,
"step": 2015,
"valid_targets_mean": 1482.8,
"valid_targets_min": 669
},
{
"epoch": 5.06265664160401,
"grad_norm": 0.8058330367764247,
"learning_rate": 2.9175453727918478e-06,
"loss": 0.3369,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3252519965171814,
"step": 2020,
"valid_targets_mean": 1552.9,
"valid_targets_min": 635
},
{
"epoch": 5.075187969924812,
"grad_norm": 0.8859191425899697,
"learning_rate": 2.8421482327204565e-06,
"loss": 0.3569,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36055076122283936,
"step": 2025,
"valid_targets_mean": 1473.4,
"valid_targets_min": 623
},
{
"epoch": 5.087719298245614,
"grad_norm": 0.8268710147356328,
"learning_rate": 2.7676635432109813e-06,
"loss": 0.3399,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3586581349372864,
"step": 2030,
"valid_targets_mean": 1494.9,
"valid_targets_min": 603
},
{
"epoch": 5.100250626566416,
"grad_norm": 0.7794366089521846,
"learning_rate": 2.6940952653418674e-06,
"loss": 0.347,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3325762152671814,
"step": 2035,
"valid_targets_mean": 1564.1,
"valid_targets_min": 502
},
{
"epoch": 5.112781954887218,
"grad_norm": 0.789354683665489,
"learning_rate": 2.6214473114569995e-06,
"loss": 0.3445,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3338083028793335,
"step": 2040,
"valid_targets_mean": 1670.4,
"valid_targets_min": 728
},
{
"epoch": 5.12531328320802,
"grad_norm": 0.87114480518287,
"learning_rate": 2.549723544957652e-06,
"loss": 0.3597,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.37028050422668457,
"step": 2045,
"valid_targets_mean": 1722.1,
"valid_targets_min": 657
},
{
"epoch": 5.137844611528822,
"grad_norm": 0.9067749222243434,
"learning_rate": 2.4789277800970247e-06,
"loss": 0.3489,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.363986611366272,
"step": 2050,
"valid_targets_mean": 1255.3,
"valid_targets_min": 622
},
{
"epoch": 5.150375939849624,
"grad_norm": 0.8809799469692904,
"learning_rate": 2.409063781777412e-06,
"loss": 0.3579,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35490089654922485,
"step": 2055,
"valid_targets_mean": 1388.8,
"valid_targets_min": 600
},
{
"epoch": 5.162907268170426,
"grad_norm": 0.8021691453837716,
"learning_rate": 2.3401352653499785e-06,
"loss": 0.349,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34150004386901855,
"step": 2060,
"valid_targets_mean": 1621.1,
"valid_targets_min": 678
},
{
"epoch": 5.175438596491228,
"grad_norm": 1.0492221755553988,
"learning_rate": 2.2721458964171704e-06,
"loss": 0.3566,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3937349021434784,
"step": 2065,
"valid_targets_mean": 1146.8,
"valid_targets_min": 649
},
{
"epoch": 5.18796992481203,
"grad_norm": 0.8387344530484291,
"learning_rate": 2.2050992906378023e-06,
"loss": 0.342,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3412122130393982,
"step": 2070,
"valid_targets_mean": 1392.2,
"valid_targets_min": 637
},
{
"epoch": 5.200501253132832,
"grad_norm": 1.0394180201727279,
"learning_rate": 2.1389990135347593e-06,
"loss": 0.3486,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.397787868976593,
"step": 2075,
"valid_targets_mean": 1076.6,
"valid_targets_min": 665
},
{
"epoch": 5.213032581453634,
"grad_norm": 0.8677433903945426,
"learning_rate": 2.073848580305382e-06,
"loss": 0.3484,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35740649700164795,
"step": 2080,
"valid_targets_mean": 1429.4,
"valid_targets_min": 768
},
{
"epoch": 5.225563909774436,
"grad_norm": 0.8209101680229015,
"learning_rate": 2.0096514556345448e-06,
"loss": 0.3383,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35316136479377747,
"step": 2085,
"valid_targets_mean": 1729.0,
"valid_targets_min": 767
},
{
"epoch": 5.238095238095238,
"grad_norm": 0.7997489355459917,
"learning_rate": 1.9464110535103885e-06,
"loss": 0.355,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33274149894714355,
"step": 2090,
"valid_targets_mean": 1570.3,
"valid_targets_min": 635
},
{
"epoch": 5.25062656641604,
"grad_norm": 0.8865845295441469,
"learning_rate": 1.8841307370427708e-06,
"loss": 0.3474,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35728734731674194,
"step": 2095,
"valid_targets_mean": 1383.3,
"valid_targets_min": 703
},
{
"epoch": 5.2631578947368425,
"grad_norm": 0.9292445555776634,
"learning_rate": 1.822813818284428e-06,
"loss": 0.3551,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3874461054801941,
"step": 2100,
"valid_targets_mean": 1317.5,
"valid_targets_min": 676
},
{
"epoch": 5.275689223057644,
"grad_norm": 0.8077993677313544,
"learning_rate": 1.76246355805481e-06,
"loss": 0.3345,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3541961908340454,
"step": 2105,
"valid_targets_mean": 1608.7,
"valid_targets_min": 580
},
{
"epoch": 5.288220551378446,
"grad_norm": 0.8473451759624002,
"learning_rate": 1.7030831657667125e-06,
"loss": 0.3548,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35867220163345337,
"step": 2110,
"valid_targets_mean": 1477.7,
"valid_targets_min": 693
},
{
"epoch": 5.3007518796992485,
"grad_norm": 0.8117634827136613,
"learning_rate": 1.6446757992555662e-06,
"loss": 0.353,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33366554975509644,
"step": 2115,
"valid_targets_mean": 1496.1,
"valid_targets_min": 688
},
{
"epoch": 5.31328320802005,
"grad_norm": 0.7969470271219807,
"learning_rate": 1.5872445646115253e-06,
"loss": 0.3608,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3518367409706116,
"step": 2120,
"valid_targets_mean": 1646.6,
"valid_targets_min": 749
},
{
"epoch": 5.325814536340852,
"grad_norm": 0.9442497476680294,
"learning_rate": 1.5307925160142767e-06,
"loss": 0.3527,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3696008324623108,
"step": 2125,
"valid_targets_mean": 1246.6,
"valid_targets_min": 618
},
{
"epoch": 5.338345864661654,
"grad_norm": 0.890713206901813,
"learning_rate": 1.4753226555706169e-06,
"loss": 0.3601,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36075615882873535,
"step": 2130,
"valid_targets_mean": 1409.6,
"valid_targets_min": 696
},
{
"epoch": 5.350877192982456,
"grad_norm": 0.7836357760961904,
"learning_rate": 1.4208379331548127e-06,
"loss": 0.344,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32944750785827637,
"step": 2135,
"valid_targets_mean": 1811.9,
"valid_targets_min": 761
},
{
"epoch": 5.363408521303258,
"grad_norm": 0.790615433618982,
"learning_rate": 1.3673412462517165e-06,
"loss": 0.347,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3488330841064453,
"step": 2140,
"valid_targets_mean": 1668.8,
"valid_targets_min": 718
},
{
"epoch": 5.37593984962406,
"grad_norm": 0.9850953675065485,
"learning_rate": 1.3148354398026753e-06,
"loss": 0.3601,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3874185085296631,
"step": 2145,
"valid_targets_mean": 1317.6,
"valid_targets_min": 784
},
{
"epoch": 5.388471177944862,
"grad_norm": 0.7871735563246215,
"learning_rate": 1.2633233060542538e-06,
"loss": 0.3378,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3537478744983673,
"step": 2150,
"valid_targets_mean": 1825.2,
"valid_targets_min": 818
},
{
"epoch": 5.401002506265664,
"grad_norm": 0.8663311136018653,
"learning_rate": 1.2128075844097321e-06,
"loss": 0.3641,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34863555431365967,
"step": 2155,
"valid_targets_mean": 1305.8,
"valid_targets_min": 613
},
{
"epoch": 5.413533834586466,
"grad_norm": 0.8444399292374044,
"learning_rate": 1.163290961283423e-06,
"loss": 0.3504,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3386199474334717,
"step": 2160,
"valid_targets_mean": 1479.9,
"valid_targets_min": 732
},
{
"epoch": 5.4260651629072685,
"grad_norm": 0.7455079817083609,
"learning_rate": 1.114776069957817e-06,
"loss": 0.3478,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3595254421234131,
"step": 2165,
"valid_targets_mean": 1820.9,
"valid_targets_min": 559
},
{
"epoch": 5.43859649122807,
"grad_norm": 0.8420039526065428,
"learning_rate": 1.0672654904435364e-06,
"loss": 0.3531,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.329376757144928,
"step": 2170,
"valid_targets_mean": 1556.8,
"valid_targets_min": 686
},
{
"epoch": 5.451127819548872,
"grad_norm": 0.798992177534451,
"learning_rate": 1.0207617493421385e-06,
"loss": 0.3302,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3616739511489868,
"step": 2175,
"valid_targets_mean": 1766.6,
"valid_targets_min": 726
},
{
"epoch": 5.4636591478696745,
"grad_norm": 0.7437629572079977,
"learning_rate": 9.752673197117456e-07,
"loss": 0.3404,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34114813804626465,
"step": 2180,
"valid_targets_mean": 1888.8,
"valid_targets_min": 800
},
{
"epoch": 5.476190476190476,
"grad_norm": 0.8454039260128821,
"learning_rate": 9.307846209355342e-07,
"loss": 0.3504,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3508765995502472,
"step": 2185,
"valid_targets_mean": 1470.9,
"valid_targets_min": 840
},
{
"epoch": 5.488721804511278,
"grad_norm": 1.0242145963904694,
"learning_rate": 8.873160185930674e-07,
"loss": 0.3701,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3790890574455261,
"step": 2190,
"valid_targets_mean": 1180.7,
"valid_targets_min": 512
},
{
"epoch": 5.50125313283208,
"grad_norm": 0.8818078968821532,
"learning_rate": 8.448638243344942e-07,
"loss": 0.3529,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3618394136428833,
"step": 2195,
"valid_targets_mean": 1515.6,
"valid_targets_min": 734
},
{
"epoch": 5.513784461152882,
"grad_norm": 1.038219375367169,
"learning_rate": 8.034302957576234e-07,
"loss": 0.364,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.37004348635673523,
"step": 2200,
"valid_targets_mean": 1204.1,
"valid_targets_min": 768
},
{
"epoch": 5.526315789473684,
"grad_norm": 0.9570498930145379,
"learning_rate": 7.63017636287855e-07,
"loss": 0.3666,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3601672649383545,
"step": 2205,
"valid_targets_mean": 1163.2,
"valid_targets_min": 527
},
{
"epoch": 5.538847117794486,
"grad_norm": 0.8361698631636353,
"learning_rate": 7.236279950610136e-07,
"loss": 0.352,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36759644746780396,
"step": 2210,
"valid_targets_mean": 1655.2,
"valid_targets_min": 576
},
{
"epoch": 5.551378446115288,
"grad_norm": 0.9285533875940273,
"learning_rate": 6.852634668090452e-07,
"loss": 0.3431,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36507898569107056,
"step": 2215,
"valid_targets_mean": 1249.8,
"valid_targets_min": 697
},
{
"epoch": 5.56390977443609,
"grad_norm": 0.9419414098830893,
"learning_rate": 6.479260917486296e-07,
"loss": 0.3498,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36119532585144043,
"step": 2220,
"valid_targets_mean": 1252.7,
"valid_targets_min": 670
},
{
"epoch": 5.576441102756892,
"grad_norm": 0.8711529899588126,
"learning_rate": 6.116178554726771e-07,
"loss": 0.3376,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3531295657157898,
"step": 2225,
"valid_targets_mean": 1507.8,
"valid_targets_min": 705
},
{
"epoch": 5.5889724310776945,
"grad_norm": 0.8284481667772925,
"learning_rate": 5.763406888447432e-07,
"loss": 0.3356,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3536621332168579,
"step": 2230,
"valid_targets_mean": 1475.7,
"valid_targets_min": 668
},
{
"epoch": 5.601503759398496,
"grad_norm": 0.8662173388424959,
"learning_rate": 5.420964678963314e-07,
"loss": 0.3533,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3492635488510132,
"step": 2235,
"valid_targets_mean": 1494.6,
"valid_targets_min": 804
},
{
"epoch": 5.614035087719298,
"grad_norm": 0.9354045755972811,
"learning_rate": 5.088870137271396e-07,
"loss": 0.374,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3668023347854614,
"step": 2240,
"valid_targets_mean": 1324.1,
"valid_targets_min": 668
},
{
"epoch": 5.6265664160401005,
"grad_norm": 0.7423454160245688,
"learning_rate": 4.767140924082059e-07,
"loss": 0.3425,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3354753255844116,
"step": 2245,
"valid_targets_mean": 1914.1,
"valid_targets_min": 751
},
{
"epoch": 5.639097744360902,
"grad_norm": 0.879259246955938,
"learning_rate": 4.4557941488799995e-07,
"loss": 0.3384,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35705995559692383,
"step": 2250,
"valid_targets_mean": 1369.6,
"valid_targets_min": 692
},
{
"epoch": 5.651629072681704,
"grad_norm": 1.0842382888757587,
"learning_rate": 4.154846369014198e-07,
"loss": 0.3424,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33172884583473206,
"step": 2255,
"valid_targets_mean": 1276.8,
"valid_targets_min": 640
},
{
"epoch": 5.664160401002507,
"grad_norm": 0.9302183552170346,
"learning_rate": 3.8643135888175145e-07,
"loss": 0.3463,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3444492816925049,
"step": 2260,
"valid_targets_mean": 1480.1,
"valid_targets_min": 718
},
{
"epoch": 5.676691729323308,
"grad_norm": 0.908184468618399,
"learning_rate": 3.5842112587555213e-07,
"loss": 0.3589,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35247802734375,
"step": 2265,
"valid_targets_mean": 1239.0,
"valid_targets_min": 670
},
{
"epoch": 5.68922305764411,
"grad_norm": 1.2712211577471397,
"learning_rate": 3.314554274604964e-07,
"loss": 0.3497,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3652925193309784,
"step": 2270,
"valid_targets_mean": 1641.1,
"valid_targets_min": 734
},
{
"epoch": 5.701754385964913,
"grad_norm": 0.7963337935187879,
"learning_rate": 3.055356976661417e-07,
"loss": 0.3449,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33966144919395447,
"step": 2275,
"valid_targets_mean": 1846.9,
"valid_targets_min": 506
},
{
"epoch": 5.714285714285714,
"grad_norm": 0.8327621927793852,
"learning_rate": 2.8066331489768894e-07,
"loss": 0.3628,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3613581657409668,
"step": 2280,
"valid_targets_mean": 1714.9,
"valid_targets_min": 654
},
{
"epoch": 5.726817042606516,
"grad_norm": 0.8144928766077396,
"learning_rate": 2.5683960186265954e-07,
"loss": 0.3453,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31915366649627686,
"step": 2285,
"valid_targets_mean": 1746.2,
"valid_targets_min": 755
},
{
"epoch": 5.739348370927319,
"grad_norm": 0.6630996787867633,
"learning_rate": 2.3406582550056455e-07,
"loss": 0.3366,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31774675846099854,
"step": 2290,
"valid_targets_mean": 2273.2,
"valid_targets_min": 908
},
{
"epoch": 5.7518796992481205,
"grad_norm": 0.8547740880071824,
"learning_rate": 2.1234319691553206e-07,
"loss": 0.3493,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3441549241542816,
"step": 2295,
"valid_targets_mean": 1516.8,
"valid_targets_min": 633
},
{
"epoch": 5.764411027568922,
"grad_norm": 0.9343549299288464,
"learning_rate": 1.9167287131188982e-07,
"loss": 0.3547,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39410120248794556,
"step": 2300,
"valid_targets_mean": 1385.6,
"valid_targets_min": 738
},
{
"epoch": 5.776942355889724,
"grad_norm": 0.7370330092462523,
"learning_rate": 1.7205594793273882e-07,
"loss": 0.3609,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3407435715198517,
"step": 2305,
"valid_targets_mean": 1888.4,
"valid_targets_min": 743
},
{
"epoch": 5.7894736842105265,
"grad_norm": 0.8021220836906139,
"learning_rate": 1.5349347000149784e-07,
"loss": 0.3576,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3627573251724243,
"step": 2310,
"valid_targets_mean": 1671.4,
"valid_targets_min": 788
},
{
"epoch": 5.802005012531328,
"grad_norm": 0.8883295497800556,
"learning_rate": 1.359864246664233e-07,
"loss": 0.3584,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3748079538345337,
"step": 2315,
"valid_targets_mean": 1306.4,
"valid_targets_min": 697
},
{
"epoch": 5.81453634085213,
"grad_norm": 0.8790652000994345,
"learning_rate": 1.19535742948107e-07,
"loss": 0.3489,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3644014596939087,
"step": 2320,
"valid_targets_mean": 1427.2,
"valid_targets_min": 640
},
{
"epoch": 5.827067669172933,
"grad_norm": 0.8377812295959398,
"learning_rate": 1.0414229968997325e-07,
"loss": 0.3518,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3724502921104431,
"step": 2325,
"valid_targets_mean": 1496.2,
"valid_targets_min": 792
},
{
"epoch": 5.839598997493734,
"grad_norm": 0.7607147698350342,
"learning_rate": 8.980691351174964e-08,
"loss": 0.3545,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3551217019557953,
"step": 2330,
"valid_targets_mean": 1877.3,
"valid_targets_min": 766
},
{
"epoch": 5.852130325814536,
"grad_norm": 0.9043120198314769,
"learning_rate": 7.65303467659373e-08,
"loss": 0.35,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3573995530605316,
"step": 2335,
"valid_targets_mean": 1316.8,
"valid_targets_min": 691
},
{
"epoch": 5.864661654135339,
"grad_norm": 0.8347596516353907,
"learning_rate": 6.431330549726555e-08,
"loss": 0.3584,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.359147846698761,
"step": 2340,
"valid_targets_mean": 1556.8,
"valid_targets_min": 707
},
{
"epoch": 5.87719298245614,
"grad_norm": 0.8303183935456322,
"learning_rate": 5.3156439405139817e-08,
"loss": 0.339,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3554539084434509,
"step": 2345,
"valid_targets_mean": 1561.7,
"valid_targets_min": 761
},
{
"epoch": 5.889724310776942,
"grad_norm": 0.8175782660251781,
"learning_rate": 4.306034180910246e-08,
"loss": 0.3503,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3494277596473694,
"step": 2350,
"valid_targets_mean": 1573.6,
"valid_targets_min": 726
},
{
"epoch": 5.902255639097744,
"grad_norm": 0.7580462701515597,
"learning_rate": 3.402554961727367e-08,
"loss": 0.3558,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3425142168998718,
"step": 2355,
"valid_targets_mean": 1730.6,
"valid_targets_min": 549
},
{
"epoch": 5.9147869674185465,
"grad_norm": 0.8314422303569725,
"learning_rate": 2.6052543297800937e-08,
"loss": 0.3559,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3679964542388916,
"step": 2360,
"valid_targets_mean": 1604.4,
"valid_targets_min": 751
},
{
"epoch": 5.927318295739348,
"grad_norm": 0.960324016846512,
"learning_rate": 1.9141746853299504e-08,
"loss": 0.3555,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3805701732635498,
"step": 2365,
"valid_targets_mean": 1361.2,
"valid_targets_min": 699
},
{
"epoch": 5.93984962406015,
"grad_norm": 0.9825903898086894,
"learning_rate": 1.3293527798317051e-08,
"loss": 0.3528,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36922788619995117,
"step": 2370,
"valid_targets_mean": 1131.9,
"valid_targets_min": 602
},
{
"epoch": 5.9523809523809526,
"grad_norm": 0.8195342765652477,
"learning_rate": 8.508197139782682e-09,
"loss": 0.3405,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3507591784000397,
"step": 2375,
"valid_targets_mean": 1598.8,
"valid_targets_min": 756
},
{
"epoch": 5.964912280701754,
"grad_norm": 0.7958920436004321,
"learning_rate": 4.786009360464583e-09,
"loss": 0.3416,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34905338287353516,
"step": 2380,
"valid_targets_mean": 1636.8,
"valid_targets_min": 722
},
{
"epoch": 5.977443609022556,
"grad_norm": 0.868391349668869,
"learning_rate": 2.127162405443084e-09,
"loss": 0.3422,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35776710510253906,
"step": 2385,
"valid_targets_mean": 1458.3,
"valid_targets_min": 573
},
{
"epoch": 5.989974937343359,
"grad_norm": 0.8621316894150931,
"learning_rate": 5.317976715790707e-10,
"loss": 0.3624,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3662847876548767,
"step": 2390,
"valid_targets_mean": 1573.8,
"valid_targets_min": 767
},
{
"epoch": 6.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33341556787490845,
"step": 2394,
"total_flos": 235219454066688.0,
"train_loss": 0.41497956364475497,
"train_runtime": 8697.8458,
"train_samples_per_second": 4.4,
"train_steps_per_second": 0.275,
"valid_targets_mean": 1707.1,
"valid_targets_min": 654
}
],
"logging_steps": 5,
"max_steps": 2394,
"num_input_tokens_seen": 0,
"num_train_epochs": 6,
"save_steps": 100,
"stateful_callbacks": {
"TrainerControl": {
"args": {
"should_epoch_stop": false,
"should_evaluate": false,
"should_log": false,
"should_save": true,
"should_training_stop": true
},
"attributes": {}
}
},
"total_flos": 235219454066688.0,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}