Upload folder using huggingface_hub

.gitattributes

@@ -68,3 +68,4 @@ checkpoints-v5.14/checkpoint-10240/eval_state.json filter=lfs diff=lfs merge=lfs
 checkpoints-v4.6++/checkpoint-2048/eval_state.json filter=lfs diff=lfs merge=lfs -text
 checkpoints-v5.14/checkpoint-21504/eval_state.json filter=lfs diff=lfs merge=lfs -text
 checkpoints-v4.6++/checkpoint-16384/eval_state.json filter=lfs diff=lfs merge=lfs -text
+checkpoints-v4.6++/checkpoint-21504/eval_state.json filter=lfs diff=lfs merge=lfs -text

checkpoints-v4.6++/checkpoint-21504/ema.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d5d096b8c611a6e9790c0f10b1062ba4dcd9c95e2acc8f772d3550c0d3e929db
+size 38935592

checkpoints-v4.6++/checkpoint-21504/eval_state.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb290c4716c5e4caa9d814ba571e8440ea98c04c9c3699973478307b07ca700a
+size 41547217

checkpoints-v4.6++/checkpoint-21504/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3da62311886c0262edebdfebd26f7fe4513b4f300d2720e4a739a636b7e39340
+size 38935624

checkpoints-v4.6++/checkpoint-21504/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14fb76f1287b42651d32b84bef3885fd35f0881fa53be2cb4af4f2d66e191b1a
+size 77992907

checkpoints-v4.6++/checkpoint-21504/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1dbb67236de71e7dfbb6f44788c059a8e56677f7151dcf072e20a638629416b7
+size 14645

checkpoints-v4.6++/checkpoint-21504/scaler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a17041ae1d9fafe7ed70338c5548d2781c247b5fe56ca8f093dfd6bb1a634cac
+size 1383

checkpoints-v4.6++/checkpoint-21504/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8d0f20f63f45b1b1775e9f97d9c6a1a2904fa04b7d5a41f30f11b9aef0479eb
+size 1465

checkpoints-v4.6++/checkpoint-21504/trainer_state.json

@@ -0,0 +1,1273 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9932104752667313,
+  "eval_steps": 1024,
+  "global_step": 21504,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.011823934229365849,
+      "grad_norm": 1.715803623199463,
+      "learning_rate": 0.000498046875,
+      "loss": 6.464813709259033,
+      "step": 256
+    },
+    {
+      "epoch": 0.023647868458731697,
+      "grad_norm": 1.6022182703018188,
+      "learning_rate": 0.000998046875,
+      "loss": 0.8224233388900757,
+      "step": 512
+    },
+    {
+      "epoch": 0.03547180268809755,
+      "grad_norm": 0.9533795714378357,
+      "learning_rate": 0.000999640996023194,
+      "loss": 0.3635375201702118,
+      "step": 768
+    },
+    {
+      "epoch": 0.047295736917463395,
+      "grad_norm": 1.1098988056182861,
+      "learning_rate": 0.0009985588674043958,
+      "loss": 0.2704107463359833,
+      "step": 1024
+    },
+    {
+      "epoch": 0.047295736917463395,
+      "eval_bleu": 0.3167128854088453,
+      "eval_ce_loss": 3.634030068846054,
+      "eval_cos_loss": 0.05990104354305627,
+      "eval_loss": 0.24878976892118584,
+      "eval_mse_loss": 0.11673666375208663,
+      "eval_rec_loss": 0.01704210073147054,
+      "flow/cos_sim": 0.9400989716727984,
+      "flow/improvement_ratio": 0.9788048792647445,
+      "flow/mag_ratio_mean": 0.9418843064678314,
+      "flow/mag_ratio_std": 0.1509957370182423,
+      "step": 1024
+    },
+    {
+      "epoch": 0.047295736917463395,
+      "eval_bleu": 0.3167128854088453,
+      "eval_ce_loss": 3.634030068846054,
+      "eval_cos_loss": 0.05990104354305627,
+      "eval_loss": 0.24878976892118584,
+      "eval_mse_loss": 0.11673666375208663,
+      "eval_rec_loss": 0.01704210073147054,
+      "eval_runtime": 142.3025,
+      "eval_samples_per_second": 196.715,
+      "eval_steps_per_second": 3.078,
+      "flow/cos_sim": 0.9400989716727984,
+      "flow/improvement_ratio": 0.9788048792647445,
+      "flow/mag_ratio_mean": 0.9418843064678314,
+      "flow/mag_ratio_std": 0.1509957370182423,
+      "step": 1024
+    },
+    {
+      "epoch": 0.05911967114682925,
+      "grad_norm": 0.6174790859222412,
+      "learning_rate": 0.0009967551747861387,
+      "loss": 0.24305246770381927,
+      "step": 1280
+    },
+    {
+      "epoch": 0.0709436053761951,
+      "grad_norm": 0.410963773727417,
+      "learning_rate": 0.000994232528651847,
+      "loss": 0.22739820182323456,
+      "step": 1536
+    },
+    {
+      "epoch": 0.08276753960556095,
+      "grad_norm": 0.6567087173461914,
+      "learning_rate": 0.0009909945800260092,
+      "loss": 0.21308675408363342,
+      "step": 1792
+    },
+    {
+      "epoch": 0.09459147383492679,
+      "grad_norm": 0.4822954833507538,
+      "learning_rate": 0.0009870460151900522,
+      "loss": 0.2054595947265625,
+      "step": 2048
+    },
+    {
+      "epoch": 0.09459147383492679,
+      "eval_bleu": 0.34012205471606577,
+      "eval_ce_loss": 3.195375862731237,
+      "eval_cos_loss": 0.05082051334349804,
+      "eval_loss": 0.20106960652761807,
+      "eval_mse_loss": 0.09635995549562315,
+      "eval_rec_loss": 0.0037663264811936274,
+      "flow/cos_sim": 0.9491795030630887,
+      "flow/improvement_ratio": 0.9779014392802704,
+      "flow/mag_ratio_mean": 0.9552222135132307,
+      "flow/mag_ratio_std": 0.10944377392802608,
+      "step": 2048
+    },
+    {
+      "epoch": 0.09459147383492679,
+      "eval_bleu": 0.34012205471606577,
+      "eval_ce_loss": 3.195375862731237,
+      "eval_cos_loss": 0.05082051334349804,
+      "eval_loss": 0.20106960652761807,
+      "eval_mse_loss": 0.09635995549562315,
+      "eval_rec_loss": 0.0037663264811936274,
+      "eval_runtime": 140.8597,
+      "eval_samples_per_second": 198.73,
+      "eval_steps_per_second": 3.109,
+      "flow/cos_sim": 0.9491795030630887,
+      "flow/improvement_ratio": 0.9779014392802704,
+      "flow/mag_ratio_mean": 0.9552222135132307,
+      "flow/mag_ratio_std": 0.10944377392802608,
+      "step": 2048
+    },
+    {
+      "epoch": 0.10641540806429264,
+      "grad_norm": 0.6277161240577698,
+      "learning_rate": 0.0009823925488998885,
+      "loss": 0.2314877063035965,
+      "step": 2304
+    },
+    {
+      "epoch": 0.1182393422936585,
+      "grad_norm": 0.5373402237892151,
+      "learning_rate": 0.0009770409161149525,
+      "loss": 0.19618386030197144,
+      "step": 2560
+    },
+    {
+      "epoch": 0.13006327652302435,
+      "grad_norm": 0.4662160277366638,
+      "learning_rate": 0.0009709988622506973,
+      "loss": 0.19010744988918304,
+      "step": 2816
+    },
+    {
+      "epoch": 0.1418872107523902,
+      "grad_norm": 0.5755115747451782,
+      "learning_rate": 0.000964275131968659,
+      "loss": 0.2221999615430832,
+      "step": 3072
+    },
+    {
+      "epoch": 0.1418872107523902,
+      "eval_bleu": 0.3548087690927611,
+      "eval_ce_loss": 3.0181641404487225,
+      "eval_cos_loss": 0.04750911592076358,
+      "eval_loss": 0.18772716694362632,
+      "eval_mse_loss": 0.09008093784695075,
+      "eval_rec_loss": 0.002350394916242269,
+      "flow/cos_sim": 0.952490894761804,
+      "flow/improvement_ratio": 0.9755349170127415,
+      "flow/mag_ratio_mean": 0.9562925890972626,
+      "flow/mag_ratio_std": 0.0839497795813296,
+      "step": 3072
+    },
+    {
+      "epoch": 0.1418872107523902,
+      "eval_bleu": 0.3548087690927611,
+      "eval_ce_loss": 3.0181641404487225,
+      "eval_cos_loss": 0.04750911592076358,
+      "eval_loss": 0.18772716694362632,
+      "eval_mse_loss": 0.09008093784695075,
+      "eval_rec_loss": 0.002350394916242269,
+      "eval_runtime": 141.8984,
+      "eval_samples_per_second": 197.275,
+      "eval_steps_per_second": 3.087,
+      "flow/cos_sim": 0.952490894761804,
+      "flow/improvement_ratio": 0.9755349170127415,
+      "flow/mag_ratio_mean": 0.9562925890972626,
+      "flow/mag_ratio_std": 0.0839497795813296,
+      "step": 3072
+    },
+    {
+      "epoch": 0.15371114498175603,
+      "grad_norm": 0.2759193181991577,
+      "learning_rate": 0.0009568794565203123,
+      "loss": 0.1921447515487671,
+      "step": 3328
+    },
+    {
+      "epoch": 0.1655350792111219,
+      "grad_norm": 0.43501198291778564,
+      "learning_rate": 0.0009488225396630347,
+      "loss": 0.18423226475715637,
+      "step": 3584
+    },
+    {
+      "epoch": 0.17735901344048774,
+      "grad_norm": 1.0013426542282104,
+      "learning_rate": 0.0009401160421685646,
+      "loss": 0.1921251118183136,
+      "step": 3840
+    },
+    {
+      "epoch": 0.18918294766985358,
+      "grad_norm": 0.5354504585266113,
+      "learning_rate": 0.0009307725649463714,
+      "loss": 0.18827098608016968,
+      "step": 4096
+    },
+    {
+      "epoch": 0.18918294766985358,
+      "eval_bleu": 0.37156652533025547,
+      "eval_ce_loss": 2.8341992467505746,
+      "eval_cos_loss": 0.044533780070030254,
+      "eval_loss": 0.1759945095674088,
+      "eval_mse_loss": 0.0851333888766428,
+      "eval_rec_loss": 0.0013817670421125197,
+      "flow/cos_sim": 0.955466238194949,
+      "flow/improvement_ratio": 0.974520222645372,
+      "flow/mag_ratio_mean": 0.950725546846651,
+      "flow/mag_ratio_std": 0.07412034801322304,
+      "step": 4096
+    },
+    {
+      "epoch": 0.18918294766985358,
+      "eval_bleu": 0.37156652533025547,
+      "eval_ce_loss": 2.8341992467505746,
+      "eval_cos_loss": 0.044533780070030254,
+      "eval_loss": 0.1759945095674088,
+      "eval_mse_loss": 0.0851333888766428,
+      "eval_rec_loss": 0.0013817670421125197,
+      "eval_runtime": 137.9409,
+      "eval_samples_per_second": 202.935,
+      "eval_steps_per_second": 3.175,
+      "flow/cos_sim": 0.955466238194949,
+      "flow/improvement_ratio": 0.974520222645372,
+      "flow/mag_ratio_mean": 0.950725546846651,
+      "flow/mag_ratio_std": 0.07412034801322304,
+      "step": 4096
+    },
+    {
+      "epoch": 0.20100688189921945,
+      "grad_norm": 0.339234322309494,
+      "learning_rate": 0.0009208056308063659,
+      "loss": 0.19387482106685638,
+      "step": 4352
+    },
+    {
+      "epoch": 0.2128308161285853,
+      "grad_norm": 0.33682018518447876,
+      "learning_rate": 0.0009102296648873445,
+      "loss": 0.17505280673503876,
+      "step": 4608
+    },
+    {
+      "epoch": 0.22465475035795113,
+      "grad_norm": 0.32164180278778076,
+      "learning_rate": 0.0008990599737794927,
+      "loss": 0.17230141162872314,
+      "step": 4864
+    },
+    {
+      "epoch": 0.236478684587317,
+      "grad_norm": 0.35368582606315613,
+      "learning_rate": 0.0008873127233711644,
+      "loss": 0.1889912486076355,
+      "step": 5120
+    },
+    {
+      "epoch": 0.236478684587317,
+      "eval_bleu": 0.39050883775928563,
+      "eval_ce_loss": 2.6735783679300247,
+      "eval_cos_loss": 0.04395430451726805,
+      "eval_loss": 0.1701030561096592,
+      "eval_mse_loss": 0.08448524126723476,
+      "eval_rec_loss": 0.0010150357346689928,
+      "flow/cos_sim": 0.9560457174092123,
+      "flow/improvement_ratio": 0.9749139539183003,
+      "flow/mag_ratio_mean": 0.9475442916835279,
+      "flow/mag_ratio_std": 0.06916524283587933,
+      "step": 5120
+    },
+    {
+      "epoch": 0.236478684587317,
+      "eval_bleu": 0.39050883775928563,
+      "eval_ce_loss": 2.6735783679300247,
+      "eval_cos_loss": 0.04395430451726805,
+      "eval_loss": 0.1701030561096592,
+      "eval_mse_loss": 0.08448524126723476,
+      "eval_rec_loss": 0.0010150357346689928,
+      "eval_runtime": 141.1359,
+      "eval_samples_per_second": 198.341,
+      "eval_steps_per_second": 3.103,
+      "flow/cos_sim": 0.9560457174092123,
+      "flow/improvement_ratio": 0.9749139539183003,
+      "flow/mag_ratio_mean": 0.9475442916835279,
+      "flow/mag_ratio_std": 0.06916524283587933,
+      "step": 5120
+    },
+    {
+      "epoch": 0.24830261881668284,
+      "grad_norm": 0.45707380771636963,
+      "learning_rate": 0.0008750049154520011,
+      "loss": 0.1717168092727661,
+      "step": 5376
+    },
+    {
+      "epoch": 0.2601265530460487,
+      "grad_norm": 0.4008166491985321,
+      "learning_rate": 0.0008621543631062487,
+      "loss": 0.192176952958107,
+      "step": 5632
+    },
+    {
+      "epoch": 0.27195048727541454,
+      "grad_norm": 0.4813244640827179,
+      "learning_rate": 0.0008487796649318904,
+      "loss": 0.16856931149959564,
+      "step": 5888
+    },
+    {
+      "epoch": 0.2837744215047804,
+      "grad_norm": 0.3855804204940796,
+      "learning_rate": 0.0008349001781229053,
+      "loss": 0.167781800031662,
+      "step": 6144
+    },
+    {
+      "epoch": 0.2837744215047804,
+      "eval_bleu": 0.3873117767737796,
+      "eval_ce_loss": 2.620973763128394,
+      "eval_cos_loss": 0.04299732028893684,
+      "eval_loss": 0.16611560702868247,
+      "eval_mse_loss": 0.08240045485465222,
+      "eval_rec_loss": 0.0007862089328456151,
+      "flow/cos_sim": 0.9570026853585352,
+      "flow/improvement_ratio": 0.9753599707122262,
+      "flow/mag_ratio_mean": 0.9637148648364359,
+      "flow/mag_ratio_std": 0.06670070579062858,
+      "step": 6144
+    },
+    {
+      "epoch": 0.2837744215047804,
+      "eval_bleu": 0.3873117767737796,
+      "eval_ce_loss": 2.620973763128394,
+      "eval_cos_loss": 0.04299732028893684,
+      "eval_loss": 0.16611560702868247,
+      "eval_mse_loss": 0.08240045485465222,
+      "eval_rec_loss": 0.0007862089328456151,
+      "eval_runtime": 140.8656,
+      "eval_samples_per_second": 198.721,
+      "eval_steps_per_second": 3.109,
+      "flow/cos_sim": 0.9570026853585352,
+      "flow/improvement_ratio": 0.9753599707122262,
+      "flow/mag_ratio_mean": 0.9637148648364359,
+      "flow/mag_ratio_std": 0.06670070579062858,
+      "step": 6144
+    },
+    {
+      "epoch": 0.2955983557341462,
+      "grad_norm": 0.4272307753562927,
+      "learning_rate": 0.0008205359904536107,
+      "loss": 0.1636379510164261,
+      "step": 6400
+    },
+    {
+      "epoch": 0.30742228996351206,
+      "grad_norm": 0.2547714114189148,
+      "learning_rate": 0.0008057078912056363,
+      "loss": 0.1644349843263626,
+      "step": 6656
+    },
+    {
+      "epoch": 0.3192462241928779,
+      "grad_norm": 0.2988753318786621,
+      "learning_rate": 0.0007904373410796086,
+      "loss": 0.1617669314146042,
+      "step": 6912
+    },
+    {
+      "epoch": 0.3310701584222438,
+      "grad_norm": 0.34734079241752625,
+      "learning_rate": 0.0007747464411350876,
+      "loss": 0.17707984149456024,
+      "step": 7168
+    },
+    {
+      "epoch": 0.3310701584222438,
+      "eval_bleu": 0.4023119380727991,
+      "eval_ce_loss": 2.5459160527137863,
+      "eval_cos_loss": 0.041424600879521406,
+      "eval_loss": 0.16110431763543387,
+      "eval_mse_loss": 0.07988655509333632,
+      "eval_rec_loss": 0.000697822997525396,
+      "flow/cos_sim": 0.9585754132434113,
+      "flow/improvement_ratio": 0.9751739552304081,
+      "flow/mag_ratio_mean": 0.9562308448090401,
+      "flow/mag_ratio_std": 0.0628269774133348,
+      "step": 7168
+    },
+    {
+      "epoch": 0.3310701584222438,
+      "eval_bleu": 0.4023119380727991,
+      "eval_ce_loss": 2.5459160527137863,
+      "eval_cos_loss": 0.041424600879521406,
+      "eval_loss": 0.16110431763543387,
+      "eval_mse_loss": 0.07988655509333632,
+      "eval_rec_loss": 0.000697822997525396,
+      "eval_runtime": 142.9617,
+      "eval_samples_per_second": 195.808,
+      "eval_steps_per_second": 3.064,
+      "flow/cos_sim": 0.9585754132434113,
+      "flow/improvement_ratio": 0.9751739552304081,
+      "flow/mag_ratio_mean": 0.9562308448090401,
+      "flow/mag_ratio_std": 0.0628269774133348,
+      "step": 7168
+    },
+    {
+      "epoch": 0.34289409265160964,
+      "grad_norm": 0.33408239483833313,
+      "learning_rate": 0.000758657900803716,
+      "loss": 0.16327416896820068,
+      "step": 7424
+    },
+    {
+      "epoch": 0.3547180268809755,
+      "grad_norm": 0.35198885202407837,
+      "learning_rate": 0.000742195005021869,
+      "loss": 0.17981402575969696,
+      "step": 7680
+    },
+    {
+      "epoch": 0.3665419611103413,
+      "grad_norm": 0.3214796483516693,
+      "learning_rate": 0.0007253815805303786,
+      "loss": 0.1649726778268814,
+      "step": 7936
+    },
+    {
+      "epoch": 0.37836589533970716,
+      "grad_norm": 0.3879843056201935,
+      "learning_rate": 0.0007082419613901028,
+      "loss": 0.16051222383975983,
+      "step": 8192
+    },
+    {
+      "epoch": 0.37836589533970716,
+      "eval_bleu": 0.4072572211905982,
+      "eval_ce_loss": 2.468050208265923,
+      "eval_cos_loss": 0.04050531620202271,
+      "eval_loss": 0.15748883212265902,
+      "eval_mse_loss": 0.07867741902873396,
+      "eval_rec_loss": 0.0007193771548227398,
+      "flow/cos_sim": 0.9594946978571208,
+      "flow/improvement_ratio": 0.9744812314096651,
+      "flow/mag_ratio_mean": 0.9578949354554964,
+      "flow/mag_ratio_std": 0.05807533837242486,
+      "step": 8192
+    },
+    {
+      "epoch": 0.37836589533970716,
+      "eval_bleu": 0.4072572211905982,
+      "eval_ce_loss": 2.468050208265923,
+      "eval_cos_loss": 0.04050531620202271,
+      "eval_loss": 0.15748883212265902,
+      "eval_mse_loss": 0.07867741902873396,
+      "eval_rec_loss": 0.0007193771548227398,
+      "eval_runtime": 143.765,
+      "eval_samples_per_second": 194.714,
+      "eval_steps_per_second": 3.047,
+      "flow/cos_sim": 0.9594946978571208,
+      "flow/improvement_ratio": 0.9744812314096651,
+      "flow/mag_ratio_mean": 0.9578949354554964,
+      "flow/mag_ratio_std": 0.05807533837242486,
+      "step": 8192
+    },
+    {
+      "epoch": 0.390189829569073,
+      "grad_norm": 0.46790069341659546,
+      "learning_rate": 0.0006908009537632514,
+      "loss": 0.17205694317817688,
+      "step": 8448
+    },
+    {
+      "epoch": 0.4020137637984389,
+      "grad_norm": 0.5188812613487244,
+      "learning_rate": 0.0006730838000114403,
+      "loss": 0.1615457683801651,
+      "step": 8704
+    },
+    {
+      "epoch": 0.41383769802780473,
+      "grad_norm": 0.2284778207540512,
+      "learning_rate": 0.0006551161421624341,
+      "loss": 0.15933068096637726,
+      "step": 8960
+    },
+    {
+      "epoch": 0.4256616322571706,
+      "grad_norm": 0.40854525566101074,
+      "learning_rate": 0.0006369239847984517,
+      "loss": 0.15647219121456146,
+      "step": 9216
+    },
+    {
+      "epoch": 0.4256616322571706,
+      "eval_bleu": 0.40657445185869284,
+      "eval_ce_loss": 2.4746508304386925,
+      "eval_cos_loss": 0.04025242353053806,
+      "eval_loss": 0.15691306173257089,
+      "eval_mse_loss": 0.07818690961286358,
+      "eval_rec_loss": 0.0004613867755349381,
+      "flow/cos_sim": 0.959747589342126,
+      "flow/improvement_ratio": 0.9742999698745606,
+      "flow/mag_ratio_mean": 0.9624219668782465,
+      "flow/mag_ratio_std": 0.05679564109097605,
+      "step": 9216
+    },
+    {
+      "epoch": 0.4256616322571706,
+      "eval_bleu": 0.40657445185869284,
+      "eval_ce_loss": 2.4746508304386925,
+      "eval_cos_loss": 0.04025242353053806,
+      "eval_loss": 0.15691306173257089,
+      "eval_mse_loss": 0.07818690961286358,
+      "eval_rec_loss": 0.0004613867755349381,
+      "eval_runtime": 142.9306,
+      "eval_samples_per_second": 195.85,
+      "eval_steps_per_second": 3.064,
+      "flow/cos_sim": 0.959747589342126,
+      "flow/improvement_ratio": 0.9742999698745606,
+      "flow/mag_ratio_mean": 0.9624219668782465,
+      "flow/mag_ratio_std": 0.05679564109097605,
+      "step": 9216
+    },
+    {
+      "epoch": 0.4374855664865364,
+      "grad_norm": 0.28328362107276917,
+      "learning_rate": 0.0006185336574197479,
+      "loss": 0.16054370999336243,
+      "step": 9472
+    },
+    {
+      "epoch": 0.44930950071590225,
+      "grad_norm": 0.41999542713165283,
+      "learning_rate": 0.0005999717763379407,
+      "loss": 0.1585461050271988,
+      "step": 9728
+    },
+    {
+      "epoch": 0.4611334349452681,
+      "grad_norm": 0.4852796494960785,
+      "learning_rate": 0.0005812652061542363,
+      "loss": 0.15610648691654205,
+      "step": 9984
+    },
+    {
+      "epoch": 0.472957369174634,
+      "grad_norm": 0.23052752017974854,
+      "learning_rate": 0.0005624410208783071,
+      "loss": 0.15434160828590393,
+      "step": 10240
+    },
+    {
+      "epoch": 0.472957369174634,
+      "eval_bleu": 0.4162788088251929,
+      "eval_ce_loss": 2.407507280508677,
+      "eval_cos_loss": 0.03962239062149911,
+      "eval_loss": 0.1538733235777241,
+      "eval_mse_loss": 0.0772547679932014,
+      "eval_rec_loss": 0.0004310996294710397,
+      "flow/cos_sim": 0.9603776199632583,
+      "flow/improvement_ratio": 0.9764508692096902,
+      "flow/mag_ratio_mean": 0.9602952395399956,
+      "flow/mag_ratio_std": 0.05461964382020306,
+      "step": 10240
+    },
+    {
+      "epoch": 0.472957369174634,
+      "eval_bleu": 0.4162788088251929,
+      "eval_ce_loss": 2.407507280508677,
+      "eval_cos_loss": 0.03962239062149911,
+      "eval_loss": 0.1538733235777241,
+      "eval_mse_loss": 0.0772547679932014,
+      "eval_rec_loss": 0.0004310996294710397,
+      "eval_runtime": 144.145,
+      "eval_samples_per_second": 194.2,
+      "eval_steps_per_second": 3.039,
+      "flow/cos_sim": 0.9603776199632583,
+      "flow/improvement_ratio": 0.9764508692096902,
+      "flow/mag_ratio_mean": 0.9602952395399956,
+      "flow/mag_ratio_std": 0.05461964382020306,
+      "step": 10240
+    },
+    {
+      "epoch": 0.48478130340399983,
+      "grad_norm": 0.2624798119068146,
+      "learning_rate": 0.0005435264647440881,
+      "loss": 0.15566085278987885,
+      "step": 10496
+    },
+    {
+      "epoch": 0.49660523763336567,
+      "grad_norm": 0.425030916929245,
+      "learning_rate": 0.000524548912779213,
+      "loss": 0.1630469262599945,
+      "step": 10752
+    },
+    {
+      "epoch": 0.5084291718627315,
+      "grad_norm": 0.4259486496448517,
+      "learning_rate": 0.0005055358311851499,
+      "loss": 0.15500524640083313,
+      "step": 11008
+    },
+    {
+      "epoch": 0.5202531060920974,
+      "grad_norm": 0.21415553987026215,
+      "learning_rate": 0.0004865147375853812,
+      "loss": 0.15434658527374268,
+      "step": 11264
+    },
+    {
+      "epoch": 0.5202531060920974,
+      "eval_bleu": 0.42080783073849026,
+      "eval_ce_loss": 2.3612416250520645,
+      "eval_cos_loss": 0.039072291262141644,
+      "eval_loss": 0.1515421306025492,
+      "eval_mse_loss": 0.07640893118600595,
+      "eval_rec_loss": 0.0003887225318220726,
+      "flow/cos_sim": 0.9609277251104241,
+      "flow/improvement_ratio": 0.9758701007387954,
+      "flow/mag_ratio_mean": 0.9595490865511437,
+      "flow/mag_ratio_std": 0.052457981330414884,
+      "step": 11264
+    },
+    {
+      "epoch": 0.5202531060920974,
+      "eval_bleu": 0.42080783073849026,
+      "eval_ce_loss": 2.3612416250520645,
+      "eval_cos_loss": 0.039072291262141644,
+      "eval_loss": 0.1515421306025492,
+      "eval_mse_loss": 0.07640893118600595,
+      "eval_rec_loss": 0.0003887225318220726,
+      "eval_runtime": 142.8047,
+      "eval_samples_per_second": 196.023,
+      "eval_steps_per_second": 3.067,
+      "flow/cos_sim": 0.9609277251104241,
+      "flow/improvement_ratio": 0.9758701007387954,
+      "flow/mag_ratio_mean": 0.9595490865511437,
+      "flow/mag_ratio_std": 0.052457981330414884,
+      "step": 11264
+    },
+    {
+      "epoch": 0.5320770403214632,
+      "grad_norm": 0.7951144576072693,
+      "learning_rate": 0.0004675131611991607,
+      "loss": 0.1532483696937561,
+      "step": 11520
+    },
+    {
+      "epoch": 0.5439009745508291,
+      "grad_norm": 0.26155829429626465,
+      "learning_rate": 0.0004485586029984899,
+      "loss": 0.1524871289730072,
+      "step": 11776
+    },
+    {
+      "epoch": 0.5557249087801949,
+      "grad_norm": 0.18861305713653564,
+      "learning_rate": 0.00042967849590597266,
+      "loss": 0.15207631886005402,
+      "step": 12032
+    },
+    {
+      "epoch": 0.5675488430095608,
+      "grad_norm": 0.2819105088710785,
+      "learning_rate": 0.0004109001650911621,
+      "loss": 0.15974190831184387,
+      "step": 12288
+    },
+    {
+      "epoch": 0.5675488430095608,
+      "eval_bleu": 0.4257705896970786,
+      "eval_ce_loss": 2.349550328842581,
+      "eval_cos_loss": 0.04070986842643044,
+      "eval_loss": 0.1541640184581552,
+      "eval_mse_loss": 0.0792681596764963,
+      "eval_rec_loss": 0.00033836385046846085,
+      "flow/cos_sim": 0.9592901502297894,
+      "flow/improvement_ratio": 0.9748820252886646,
+      "flow/mag_ratio_mean": 0.956312407777734,
+      "flow/mag_ratio_std": 0.05244758199543169,
+      "step": 12288
+    },
+    {
+      "epoch": 0.5675488430095608,
+      "eval_bleu": 0.4257705896970786,
+      "eval_ce_loss": 2.349550328842581,
+      "eval_cos_loss": 0.04070986842643044,
+      "eval_loss": 0.1541640184581552,
+      "eval_mse_loss": 0.0792681596764963,
+      "eval_rec_loss": 0.00033836385046846085,
+      "eval_runtime": 144.6517,
+      "eval_samples_per_second": 193.52,
+      "eval_steps_per_second": 3.028,
+      "flow/cos_sim": 0.9592901502297894,
+      "flow/improvement_ratio": 0.9748820252886646,
+      "flow/mag_ratio_mean": 0.956312407777734,
+      "flow/mag_ratio_std": 0.05244758199543169,
+      "step": 12288
+    },
+    {
+      "epoch": 0.5793727772389267,
+      "grad_norm": 0.7697474360466003,
+      "learning_rate": 0.0003922507884228551,
+      "loss": 0.15206775069236755,
+      "step": 12544
+    },
+    {
+      "epoch": 0.5911967114682924,
+      "grad_norm": 0.1449330896139145,
+      "learning_rate": 0.00037375735713457723,
+      "loss": 0.1501483917236328,
+      "step": 12800
+    },
+    {
+      "epoch": 0.6030206456976583,
+      "grad_norm": 0.17718757688999176,
+      "learning_rate": 0.00035544663676018276,
+      "loss": 0.15162093937397003,
+      "step": 13056
+    },
+    {
+      "epoch": 0.6148445799270241,
+      "grad_norm": 0.2246871143579483,
+      "learning_rate": 0.00033734512839611255,
+      "loss": 0.14962096512317657,
+      "step": 13312
+    },
+    {
+      "epoch": 0.6148445799270241,
+      "eval_bleu": 0.4292785120885617,
+      "eval_ce_loss": 2.3244524306902603,
+      "eval_cos_loss": 0.03869187967537065,
+      "eval_loss": 0.14974690237205868,
+      "eval_mse_loss": 0.07580797687240931,
+      "eval_rec_loss": 0.00033616627441828415,
+      "flow/cos_sim": 0.9613081395081734,
+      "flow/improvement_ratio": 0.9738387182422968,
+      "flow/mag_ratio_mean": 0.9639241397380829,
+      "flow/mag_ratio_std": 0.05269840318012183,
+      "step": 13312
+    },
+    {
+      "epoch": 0.6148445799270241,
+      "eval_bleu": 0.4292785120885617,
+      "eval_ce_loss": 2.3244524306902603,
+      "eval_cos_loss": 0.03869187967537065,
+      "eval_loss": 0.14974690237205868,
+      "eval_mse_loss": 0.07580797687240931,
+      "eval_rec_loss": 0.00033616627441828415,
+      "eval_runtime": 143.4574,
+      "eval_samples_per_second": 195.131,
+      "eval_steps_per_second": 3.053,
+      "flow/cos_sim": 0.9613081395081734,
+      "flow/improvement_ratio": 0.9738387182422968,
+      "flow/mag_ratio_mean": 0.9639241397380829,
+      "flow/mag_ratio_std": 0.05269840318012183,
+      "step": 13312
+    },
+    {
+      "epoch": 0.62666851415639,
+      "grad_norm": 0.22714389860630035,
+      "learning_rate": 0.0003194790303463687,
+      "loss": 0.14983858168125153,
+      "step": 13568
+    },
+    {
+      "epoch": 0.6384924483857558,
+      "grad_norm": 0.2802693545818329,
+      "learning_rate": 0.00030187420020572406,
+      "loss": 0.1504410058259964,
+      "step": 13824
+    },
+    {
+      "epoch": 0.6503163826151217,
+      "grad_norm": 0.15414872765541077,
+      "learning_rate": 0.00028455611743603626,
+      "loss": 0.14932161569595337,
+      "step": 14080
+    },
+    {
+      "epoch": 0.6621403168444876,
+      "grad_norm": 0.13487546145915985,
+      "learning_rate": 0.0002675498464898373,
+      "loss": 0.14990688860416412,
+      "step": 14336
+    },
+    {
+      "epoch": 0.6621403168444876,
+      "eval_bleu": 0.4349671845891111,
+      "eval_ce_loss": 2.2670639056593314,
+      "eval_cos_loss": 0.03825397921734613,
+      "eval_loss": 0.14740464028362268,
+      "eval_mse_loss": 0.075232593793439,
+      "eval_rec_loss": 0.00033473305019092074,
+      "flow/cos_sim": 0.9617460425858084,
+      "flow/improvement_ratio": 0.9755375824834658,
+      "flow/mag_ratio_mean": 0.9641721471803918,
+      "flow/mag_ratio_std": 0.05227890214351214,
+      "step": 14336
+    },
+    {
+      "epoch": 0.6621403168444876,
+      "eval_bleu": 0.4349671845891111,
+      "eval_ce_loss": 2.2670639056593314,
+      "eval_cos_loss": 0.03825397921734613,
+      "eval_loss": 0.14740464028362268,
+      "eval_mse_loss": 0.075232593793439,
+      "eval_rec_loss": 0.00033473305019092074,
+      "eval_runtime": 142.2933,
+      "eval_samples_per_second": 196.727,
+      "eval_steps_per_second": 3.078,
+      "flow/cos_sim": 0.9617460425858084,
+      "flow/improvement_ratio": 0.9755375824834658,
+      "flow/mag_ratio_mean": 0.9641721471803918,
+      "flow/mag_ratio_std": 0.05227890214351214,
+      "step": 14336
+    },
+    {
+      "epoch": 0.6739642510738534,
+      "grad_norm": 0.6694812774658203,
+      "learning_rate": 0.0002508800005345623,
+      "loss": 0.14883804321289062,
+      "step": 14592
+    },
+    {
+      "epoch": 0.6857881853032193,
+      "grad_norm": 0.665196418762207,
+      "learning_rate": 0.00023457070582992562,
+      "loss": 0.1497967392206192,
+      "step": 14848
+    },
+    {
+      "epoch": 0.6976121195325851,
+      "grad_norm": 0.7369906306266785,
+      "learning_rate": 0.00021864556680999692,
+      "loss": 0.14808616042137146,
+      "step": 15104
+    },
+    {
+      "epoch": 0.709436053761951,
+      "grad_norm": 0.2977660298347473,
+      "learning_rate": 0.0002031276319205152,
+      "loss": 0.14767387509346008,
+      "step": 15360
+    },
+    {
+      "epoch": 0.709436053761951,
+      "eval_bleu": 0.4328262745636852,
+      "eval_ce_loss": 2.282826251907436,
+      "eval_cos_loss": 0.037827753053646385,
+      "eval_loss": 0.14708495522810988,
+      "eval_mse_loss": 0.07452186333126129,
+      "eval_rec_loss": 0.0002955303820164116,
+      "flow/cos_sim": 0.9621722616561471,
+      "flow/improvement_ratio": 0.9745410205838887,
+      "flow/mag_ratio_mean": 0.9601432559969219,
+      "flow/mag_ratio_std": 0.05038045190240695,
+      "step": 15360
+    },
+    {
+      "epoch": 0.709436053761951,
+      "eval_bleu": 0.4328262745636852,
+      "eval_ce_loss": 2.282826251907436,
+      "eval_cos_loss": 0.037827753053646385,
+      "eval_loss": 0.14708495522810988,
+      "eval_mse_loss": 0.07452186333126129,
+      "eval_rec_loss": 0.0002955303820164116,
+      "eval_runtime": 142.9403,
+      "eval_samples_per_second": 195.837,
+      "eval_steps_per_second": 3.064,
+      "flow/cos_sim": 0.9621722616561471,
+      "flow/improvement_ratio": 0.9745410205838887,
+      "flow/mag_ratio_mean": 0.9601432559969219,
+      "flow/mag_ratio_std": 0.05038045190240695,
+      "step": 15360
+    },
+    {
+      "epoch": 0.7212599879913169,
+      "grad_norm": 0.16127941012382507,
+      "learning_rate": 0.00018803936026088542,
+      "loss": 0.1476622074842453,
+      "step": 15616
+    },
+    {
+      "epoch": 0.7330839222206826,
+      "grad_norm": 0.21558411419391632,
+      "learning_rate": 0.00017340258907913464,
+      "loss": 0.14784899353981018,
+      "step": 15872
+    },
+    {
+      "epoch": 0.7449078564500485,
+      "grad_norm": 0.7104827165603638,
+      "learning_rate": 0.0001592385021668743,
+      "loss": 0.14831793308258057,
+      "step": 16128
+    },
+    {
+      "epoch": 0.7567317906794143,
+      "grad_norm": 0.14697256684303284,
+      "learning_rate": 0.0001455675992000087,
+      "loss": 0.14772531390190125,
+      "step": 16384
+    },
+    {
+      "epoch": 0.7567317906794143,
+      "eval_bleu": 0.43514709071220475,
+      "eval_ce_loss": 2.2809920245653963,
+      "eval_cos_loss": 0.037835451166175274,
+      "eval_loss": 0.1468935562935594,
+      "eval_mse_loss": 0.07440643887711715,
+      "eval_rec_loss": 0.00027381275814652794,
+      "flow/cos_sim": 0.9621645644919513,
+      "flow/improvement_ratio": 0.9732955764962113,
+      "flow/mag_ratio_mean": 0.9549076063175724,
+      "flow/mag_ratio_std": 0.050503925041861186,
+      "step": 16384
+    },
+    {
+      "epoch": 0.7567317906794143,
+      "eval_bleu": 0.43514709071220475,
+      "eval_ce_loss": 2.2809920245653963,
+      "eval_cos_loss": 0.037835451166175274,
+      "eval_loss": 0.1468935562935594,
+      "eval_mse_loss": 0.07440643887711715,
+      "eval_rec_loss": 0.00027381275814652794,
+      "eval_runtime": 143.5278,
+      "eval_samples_per_second": 195.035,
+      "eval_steps_per_second": 3.052,
+      "flow/cos_sim": 0.9621645644919513,
+      "flow/improvement_ratio": 0.9732955764962113,
+      "flow/mag_ratio_mean": 0.9549076063175724,
+      "flow/mag_ratio_std": 0.050503925041861186,
+      "step": 16384
+    },
+    {
+      "epoch": 0.7685557249087802,
+      "grad_norm": 0.16373251378536224,
+      "learning_rate": 0.000132409666069565,
+      "loss": 0.14727428555488586,
+      "step": 16640
+    },
+    {
+      "epoch": 0.780379659138146,
+      "grad_norm": 0.27290403842926025,
+      "learning_rate": 0.0001197837462455823,
+      "loss": 0.14652089774608612,
+      "step": 16896
+    },
+    {
+      "epoch": 0.7922035933675119,
+      "grad_norm": 0.16963687539100647,
+      "learning_rate": 0.00010770811321550749,
+      "loss": 0.14595653116703033,
+      "step": 17152
+    },
+    {
+      "epoch": 0.8040275275968778,
+      "grad_norm": 0.1278214007616043,
+      "learning_rate": 9.620024403698591e-05,
+      "loss": 0.14666643738746643,
+      "step": 17408
+    },
+    {
+      "epoch": 0.8040275275968778,
+      "eval_bleu": 0.43458046273658474,
+      "eval_ce_loss": 2.2552077604211083,
+      "eval_cos_loss": 0.03711736504600881,
+      "eval_loss": 0.14493908716119044,
+      "eval_mse_loss": 0.07331650395405619,
+      "eval_rec_loss": 0.00025461489209715495,
+      "flow/cos_sim": 0.9628826502251299,
+      "flow/improvement_ratio": 0.9733623802389728,
+      "flow/mag_ratio_mean": 0.96390518233112,
+      "flow/mag_ratio_std": 0.04803763700810742,
+      "step": 17408
+    },
+    {
+      "epoch": 0.8040275275968778,
+      "eval_bleu": 0.43458046273658474,
+      "eval_ce_loss": 2.2552077604211083,
+      "eval_cos_loss": 0.03711736504600881,
+      "eval_loss": 0.14493908716119044,
+      "eval_mse_loss": 0.07331650395405619,
+      "eval_rec_loss": 0.00025461489209715495,
+      "eval_runtime": 145.1612,
+      "eval_samples_per_second": 192.841,
+      "eval_steps_per_second": 3.017,
+      "flow/cos_sim": 0.9628826502251299,
+      "flow/improvement_ratio": 0.9733623802389728,
+      "flow/mag_ratio_mean": 0.96390518233112,
+      "flow/mag_ratio_std": 0.04803763700810742,
+      "step": 17408
+    },
+    {
+      "epoch": 0.8158514618262436,
+      "grad_norm": 0.4657216966152191,
+      "learning_rate": 8.527679404332429e-05,
+      "loss": 0.14539453387260437,
+      "step": 17664
+    },
+    {
+      "epoch": 0.8276753960556095,
+      "grad_norm": 0.10805023461580276,
+      "learning_rate": 7.495357273823544e-05,
+      "loss": 0.1464768648147583,
+      "step": 17920
+    },
+    {
+      "epoch": 0.8394993302849753,
+      "grad_norm": 0.11020675301551819,
+      "learning_rate": 6.524552091475183e-05,
+      "loss": 0.14626719057559967,
+      "step": 18176
+    },
+    {
+      "epoch": 0.8513232645143411,
+      "grad_norm": 0.08271762728691101,
+      "learning_rate": 5.6166689031422024e-05,
+      "loss": 0.1454443633556366,
+      "step": 18432
+    },
+    {
+      "epoch": 0.8513232645143411,
+      "eval_bleu": 0.4369459790610845,
+      "eval_ce_loss": 2.2421654413824212,
+      "eval_cos_loss": 0.03716503064994398,
+      "eval_loss": 0.14462064531499938,
+      "eval_mse_loss": 0.07339327158108694,
+      "eval_rec_loss": 0.00024590859372757766,
+      "flow/cos_sim": 0.962834993453875,
+      "flow/improvement_ratio": 0.9746005626301787,
+      "flow/mag_ratio_mean": 0.9625570138567658,
+      "flow/mag_ratio_std": 0.04861913955681128,
+      "step": 18432
+    },
+    {
+      "epoch": 0.8513232645143411,
+      "eval_bleu": 0.4369459790610845,
+      "eval_ce_loss": 2.2421654413824212,
+      "eval_cos_loss": 0.03716503064994398,
+      "eval_loss": 0.14462064531499938,
+      "eval_mse_loss": 0.07339327158108694,
+      "eval_rec_loss": 0.00024590859372757766,
+      "eval_runtime": 141.9029,
+      "eval_samples_per_second": 197.269,
+      "eval_steps_per_second": 3.087,
+      "flow/cos_sim": 0.962834993453875,
+      "flow/improvement_ratio": 0.9746005626301787,
+      "flow/mag_ratio_mean": 0.9625570138567658,
+      "flow/mag_ratio_std": 0.04861913955681128,
+      "step": 18432
+    },
+    {
+      "epoch": 0.8631471987437069,
+      "grad_norm": 0.08390416204929352,
+      "learning_rate": 4.773021687709067e-05,
+      "loss": 0.14536336064338684,
+      "step": 18688
+    },
+    {
+      "epoch": 0.8749711329730728,
+      "grad_norm": 0.10406672209501266,
+      "learning_rate": 3.994831455368719e-05,
+      "loss": 0.14549703896045685,
+      "step": 18944
+    },
+    {
+      "epoch": 0.8867950672024387,
+      "grad_norm": 0.11962409317493439,
+      "learning_rate": 3.283224480455282e-05,
+      "loss": 0.14519065618515015,
+      "step": 19200
+    },
+    {
+      "epoch": 0.8986190014318045,
+      "grad_norm": 0.9476587772369385,
+      "learning_rate": 2.639230671387627e-05,
+      "loss": 0.14600209891796112,
+      "step": 19456
+    },
+    {
+      "epoch": 0.8986190014318045,
+      "eval_bleu": 0.4387582306085107,
+      "eval_ce_loss": 2.233657001088199,
+      "eval_cos_loss": 0.03726340171596095,
+      "eval_loss": 0.14455011328809883,
+      "eval_mse_loss": 0.07358088907381716,
+      "eval_rec_loss": 0.00023317563165170985,
+      "flow/cos_sim": 0.9627366185732628,
+      "flow/improvement_ratio": 0.9742960237204756,
+      "flow/mag_ratio_mean": 0.9631893065966428,
+      "flow/mag_ratio_std": 0.04909244309005128,
+      "step": 19456
+    },
+    {
+      "epoch": 0.8986190014318045,
+      "eval_bleu": 0.4387582306085107,
+      "eval_ce_loss": 2.233657001088199,
+      "eval_cos_loss": 0.03726340171596095,
+      "eval_loss": 0.14455011328809883,
+      "eval_mse_loss": 0.07358088907381716,
+      "eval_rec_loss": 0.00023317563165170985,
+      "eval_runtime": 142.7597,
+      "eval_samples_per_second": 196.085,
+      "eval_steps_per_second": 3.068,
+      "flow/cos_sim": 0.9627366185732628,
+      "flow/improvement_ratio": 0.9742960237204756,
+      "flow/mag_ratio_mean": 0.9631893065966428,
+      "flow/mag_ratio_std": 0.04909244309005128,
+      "step": 19456
+    },
+    {
+      "epoch": 0.9104429356611704,
+      "grad_norm": 0.07215014100074768,
+      "learning_rate": 2.063782080083576e-05,
+      "loss": 0.14549361169338226,
+      "step": 19712
+    },
+    {
+      "epoch": 0.9222668698905362,
+      "grad_norm": 0.09603725373744965,
+      "learning_rate": 1.557711553001523e-05,
+      "loss": 0.14520247280597687,
+      "step": 19968
+    },
+    {
+      "epoch": 0.9340908041199021,
+      "grad_norm": 0.09211431443691254,
+      "learning_rate": 1.1217515257622269e-05,
+      "loss": 0.14462901651859283,
+      "step": 20224
+    },
+    {
+      "epoch": 0.945914738349268,
+      "grad_norm": 0.1264684647321701,
+      "learning_rate": 7.565329630950746e-06,
+      "loss": 0.14502722024917603,
+      "step": 20480
+    },
+    {
+      "epoch": 0.945914738349268,
+      "eval_bleu": 0.43821577509576853,
+      "eval_ce_loss": 2.242395851437904,
+      "eval_cos_loss": 0.036767141612833495,
+      "eval_loss": 0.14385633441070989,
+      "eval_mse_loss": 0.07265583999012704,
+      "eval_rec_loss": 0.0002519065699029499,
+      "flow/cos_sim": 0.963232872965129,
+      "flow/improvement_ratio": 0.9755431610971825,
+      "flow/mag_ratio_mean": 0.9630188521457045,
+      "flow/mag_ratio_std": 0.04812579758499311,
+      "step": 20480
+    },
+    {
+      "epoch": 0.945914738349268,
+      "eval_bleu": 0.43821577509576853,
+      "eval_ce_loss": 2.242395851437904,
+      "eval_cos_loss": 0.036767141612833495,
+      "eval_loss": 0.14385633441070989,
+      "eval_mse_loss": 0.07265583999012704,
+      "eval_rec_loss": 0.0002519065699029499,
+      "eval_runtime": 142.1306,
+      "eval_samples_per_second": 196.953,
+      "eval_steps_per_second": 3.082,
+      "flow/cos_sim": 0.963232872965129,
+      "flow/improvement_ratio": 0.9755431610971825,
+      "flow/mag_ratio_mean": 0.9630188521457045,
+      "flow/mag_ratio_std": 0.04812579758499311,
+      "step": 20480
+    },
+    {
+      "epoch": 0.9577386725786338,
+      "grad_norm": 1.0626317262649536,
+      "learning_rate": 4.62584445643166e-06,
+      "loss": 0.14507462084293365,
+      "step": 20736
+    },
+    {
+      "epoch": 0.9695626068079997,
+      "grad_norm": 0.07459881901741028,
+      "learning_rate": 2.40331404948807e-06,
+      "loss": 0.1449519246816635,
+      "step": 20992
+    },
+    {
+      "epoch": 0.9813865410373654,
+      "grad_norm": 0.29215162992477417,
+      "learning_rate": 9.009550772663965e-07,
+      "loss": 0.1453726589679718,
+      "step": 21248
+    },
+    {
+      "epoch": 0.9932104752667313,
+      "grad_norm": 0.07788247615098953,
+      "learning_rate": 1.2094190315575791e-07,
+      "loss": 0.14389921724796295,
+      "step": 21504
+    },
+    {
+      "epoch": 0.9932104752667313,
+      "eval_bleu": 0.43990490485065636,
+      "eval_ce_loss": 2.2338819160853345,
+      "eval_cos_loss": 0.03676586426352257,
+      "eval_loss": 0.1436361019673957,
+      "eval_mse_loss": 0.07269612861346436,
+      "eval_rec_loss": 0.0002469304909941062,
+      "flow/cos_sim": 0.963234153104155,
+      "flow/improvement_ratio": 0.9737974429783756,
+      "flow/mag_ratio_mean": 0.9629674149702673,
+      "flow/mag_ratio_std": 0.04823799865090684,
+      "step": 21504
+    },
+    {
+      "epoch": 0.9932104752667313,
+      "eval_bleu": 0.43990490485065636,
+      "eval_ce_loss": 2.2338819160853345,
+      "eval_cos_loss": 0.03676586426352257,
+      "eval_loss": 0.1436361019673957,
+      "eval_mse_loss": 0.07269612861346436,
+      "eval_rec_loss": 0.0002469304909941062,
+      "eval_runtime": 143.4117,
+      "eval_samples_per_second": 195.193,
+      "eval_steps_per_second": 3.054,
+      "flow/cos_sim": 0.963234153104155,
+      "flow/improvement_ratio": 0.9737974429783756,
+      "flow/mag_ratio_mean": 0.9629674149702673,
+      "flow/mag_ratio_std": 0.04823799865090684,
+      "step": 21504
+    }
+  ],
+  "logging_steps": 256,
+  "max_steps": 21651,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 1024,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 64,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoints-v4.6++/checkpoint-21504/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61a16bb839f687414b8e48611327c4b9cfddeefe38c031ca70808f9a97c476b7
+size 5137