{
  "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": 0.13531190156936646,
      "learning_rate": 0.0002490234375,
      "loss": 1.4445146322250366,
      "step": 256
    },
    {
      "epoch": 0.023647868458731697,
      "grad_norm": 0.1410810798406601,
      "learning_rate": 0.0004990234375,
      "loss": 1.3425910472869873,
      "step": 512
    },
    {
      "epoch": 0.03547180268809755,
      "grad_norm": 0.11418232321739197,
      "learning_rate": 0.000499820498011597,
      "loss": 1.3074675798416138,
      "step": 768
    },
    {
      "epoch": 0.047295736917463395,
      "grad_norm": 0.1378929316997528,
      "learning_rate": 0.0004992794337021979,
      "loss": 1.2896003723144531,
      "step": 1024
    },
    {
      "epoch": 0.047295736917463395,
      "eval_cos_loss": 0.44286489105660076,
      "eval_loss": 1.221002193074248,
      "eval_mse_loss": 1.221002193074248,
      "flow/cos_sim": 0.5571351273146938,
      "flow/improvement_ratio": 0.9647975193855425,
      "flow/mag_ratio_mean": 0.5350822540314775,
      "flow/mag_ratio_std": 0.24492420766451586,
      "step": 1024
    },
    {
      "epoch": 0.047295736917463395,
      "eval_cos_loss": 0.44286489105660076,
      "eval_loss": 1.221002193074248,
      "eval_mse_loss": 1.221002193074248,
      "eval_runtime": 37.6256,
      "eval_samples_per_second": 743.989,
      "eval_steps_per_second": 11.641,
      "flow/cos_sim": 0.5571351273146938,
      "flow/improvement_ratio": 0.9647975193855425,
      "flow/mag_ratio_mean": 0.5350822540314775,
      "flow/mag_ratio_std": 0.24492420766451586,
      "step": 1024
    },
    {
      "epoch": 0.05911967114682925,
      "grad_norm": 0.16314451396465302,
      "learning_rate": 0.0004983775873930693,
      "loss": 1.2742286920547485,
      "step": 1280
    },
    {
      "epoch": 0.0709436053761951,
      "grad_norm": 0.1408475935459137,
      "learning_rate": 0.0004971162643259235,
      "loss": 1.2636311054229736,
      "step": 1536
    },
    {
      "epoch": 0.08276753960556095,
      "grad_norm": 0.4067126512527466,
      "learning_rate": 0.0004954972900130046,
      "loss": 1.2552720308303833,
      "step": 1792
    },
    {
      "epoch": 0.09459147383492679,
      "grad_norm": 0.11263003200292587,
      "learning_rate": 0.0004935230075950261,
      "loss": 1.250608205795288,
      "step": 2048
    },
    {
      "epoch": 0.09459147383492679,
      "eval_cos_loss": 0.4280974593472807,
      "eval_loss": 1.1850101215654312,
      "eval_mse_loss": 1.1850101215654312,
      "flow/cos_sim": 0.5719025665766573,
      "flow/improvement_ratio": 0.9701844343583877,
      "flow/mag_ratio_mean": 0.5505562974437731,
      "flow/mag_ratio_std": 0.2407817519078516,
      "step": 2048
    },
    {
      "epoch": 0.09459147383492679,
      "eval_cos_loss": 0.4280974593472807,
      "eval_loss": 1.1850101215654312,
      "eval_mse_loss": 1.1850101215654312,
      "eval_runtime": 36.5824,
      "eval_samples_per_second": 765.204,
      "eval_steps_per_second": 11.973,
      "flow/cos_sim": 0.5719025665766573,
      "flow/improvement_ratio": 0.9701844343583877,
      "flow/mag_ratio_mean": 0.5505562974437731,
      "flow/mag_ratio_std": 0.2407817519078516,
      "step": 2048
    },
    {
      "epoch": 0.10641540806429264,
      "grad_norm": 0.23430709540843964,
      "learning_rate": 0.0004911962744499443,
      "loss": 1.2465660572052002,
      "step": 2304
    },
    {
      "epoch": 0.1182393422936585,
      "grad_norm": 0.12819728255271912,
      "learning_rate": 0.0004885204580574763,
      "loss": 1.2422897815704346,
      "step": 2560
    },
    {
      "epoch": 0.13006327652302435,
      "grad_norm": 0.1120893657207489,
      "learning_rate": 0.00048549943112534866,
      "loss": 1.2349907159805298,
      "step": 2816
    },
    {
      "epoch": 0.1418872107523902,
      "grad_norm": 0.12881016731262207,
      "learning_rate": 0.0004821375659843295,
      "loss": 1.2320677042007446,
      "step": 3072
    },
    {
      "epoch": 0.1418872107523902,
      "eval_cos_loss": 0.4224365182799291,
      "eval_loss": 1.1717089387379824,
      "eval_mse_loss": 1.1717089387379824,
      "flow/cos_sim": 0.5775635075079252,
      "flow/improvement_ratio": 0.9726638531303842,
      "flow/mag_ratio_mean": 0.5545849762006437,
      "flow/mag_ratio_std": 0.2390297385927749,
      "step": 3072
    },
    {
      "epoch": 0.1418872107523902,
      "eval_cos_loss": 0.4224365182799291,
      "eval_loss": 1.1717089387379824,
      "eval_mse_loss": 1.1717089387379824,
      "eval_runtime": 36.5841,
      "eval_samples_per_second": 765.167,
      "eval_steps_per_second": 11.972,
      "flow/cos_sim": 0.5775635075079252,
      "flow/improvement_ratio": 0.9726638531303842,
      "flow/mag_ratio_mean": 0.5545849762006437,
      "flow/mag_ratio_std": 0.2390297385927749,
      "step": 3072
    },
    {
      "epoch": 0.15371114498175603,
      "grad_norm": 0.14192286133766174,
      "learning_rate": 0.00047843972826015616,
      "loss": 1.229474663734436,
      "step": 3328
    },
    {
      "epoch": 0.1655350792111219,
      "grad_norm": 0.1288514882326126,
      "learning_rate": 0.00047441126983151737,
      "loss": 1.2259553670883179,
      "step": 3584
    },
    {
      "epoch": 0.17735901344048774,
      "grad_norm": 0.1571502685546875,
      "learning_rate": 0.0004700580210842823,
      "loss": 1.2254003286361694,
      "step": 3840
    },
    {
      "epoch": 0.18918294766985358,
      "grad_norm": 0.18948955833911896,
      "learning_rate": 0.0004653862824731857,
      "loss": 1.2214388847351074,
      "step": 4096
    },
    {
      "epoch": 0.18918294766985358,
      "eval_cos_loss": 0.4189364165748091,
      "eval_loss": 1.162348562179635,
      "eval_mse_loss": 1.162348562179635,
      "flow/cos_sim": 0.5810636061511628,
      "flow/improvement_ratio": 0.971539740420912,
      "flow/mag_ratio_mean": 0.5564619683783892,
      "flow/mag_ratio_std": 0.23926625594700854,
      "step": 4096
    },
    {
      "epoch": 0.18918294766985358,
      "eval_cos_loss": 0.4189364165748091,
      "eval_loss": 1.162348562179635,
      "eval_mse_loss": 1.162348562179635,
      "eval_runtime": 36.6574,
      "eval_samples_per_second": 763.638,
      "eval_steps_per_second": 11.948,
      "flow/cos_sim": 0.5810636061511628,
      "flow/improvement_ratio": 0.971539740420912,
      "flow/mag_ratio_mean": 0.5564619683783892,
      "flow/mag_ratio_std": 0.23926625594700854,
      "step": 4096
    },
    {
      "epoch": 0.20100688189921945,
      "grad_norm": 0.11497154086828232,
      "learning_rate": 0.00046040281540318296,
      "loss": 1.2192964553833008,
      "step": 4352
    },
    {
      "epoch": 0.2128308161285853,
      "grad_norm": 0.17087024450302124,
      "learning_rate": 0.00045511483244367227,
      "loss": 1.2178804874420166,
      "step": 4608
    },
    {
      "epoch": 0.22465475035795113,
      "grad_norm": 0.15084773302078247,
      "learning_rate": 0.00044952998688974635,
      "loss": 1.2137683629989624,
      "step": 4864
    },
    {
      "epoch": 0.236478684587317,
      "grad_norm": 0.17674025893211365,
      "learning_rate": 0.0004436563616855822,
      "loss": 1.214182734489441,
      "step": 5120
    },
    {
      "epoch": 0.236478684587317,
      "eval_cos_loss": 0.41552880569680095,
      "eval_loss": 1.1498510279611909,
      "eval_mse_loss": 1.1498510279611909,
      "flow/cos_sim": 0.5844712080476491,
      "flow/improvement_ratio": 0.9718176997415552,
      "flow/mag_ratio_mean": 0.5639459366939928,
      "flow/mag_ratio_std": 0.24295896026369643,
      "step": 5120
    },
    {
      "epoch": 0.236478684587317,
      "eval_cos_loss": 0.41552880569680095,
      "eval_loss": 1.1498510279611909,
      "eval_mse_loss": 1.1498510279611909,
      "eval_runtime": 36.5045,
      "eval_samples_per_second": 766.837,
      "eval_steps_per_second": 11.999,
      "flow/cos_sim": 0.5844712080476491,
      "flow/improvement_ratio": 0.9718176997415552,
      "flow/mag_ratio_mean": 0.5639459366939928,
      "flow/mag_ratio_std": 0.24295896026369643,
      "step": 5120
    },
    {
      "epoch": 0.24830261881668284,
      "grad_norm": 0.19734610617160797,
      "learning_rate": 0.00043750245772600053,
      "loss": 1.2093557119369507,
      "step": 5376
    },
    {
      "epoch": 0.2601265530460487,
      "grad_norm": 0.11800088733434677,
      "learning_rate": 0.00043107718155312435,
      "loss": 1.2076833248138428,
      "step": 5632
    },
    {
      "epoch": 0.27195048727541454,
      "grad_norm": 0.11836104094982147,
      "learning_rate": 0.0004243898324659452,
      "loss": 1.2095065116882324,
      "step": 5888
    },
    {
      "epoch": 0.2837744215047804,
      "grad_norm": 0.14233353734016418,
      "learning_rate": 0.00041745008906145264,
      "loss": 1.2082980871200562,
      "step": 6144
    },
    {
      "epoch": 0.2837744215047804,
      "eval_cos_loss": 0.41333945823586693,
      "eval_loss": 1.1455941828962874,
      "eval_mse_loss": 1.1455941828962874,
      "flow/cos_sim": 0.5866605730633757,
      "flow/improvement_ratio": 0.9727034600085864,
      "flow/mag_ratio_mean": 0.5605599706031416,
      "flow/mag_ratio_std": 0.24318457544530364,
      "step": 6144
    },
    {
      "epoch": 0.2837744215047804,
      "eval_cos_loss": 0.41333945823586693,
      "eval_loss": 1.1455941828962874,
      "eval_mse_loss": 1.1455941828962874,
      "eval_runtime": 36.3926,
      "eval_samples_per_second": 769.196,
      "eval_steps_per_second": 12.035,
      "flow/cos_sim": 0.5866605730633757,
      "flow/improvement_ratio": 0.9727034600085864,
      "flow/mag_ratio_mean": 0.5605599706031416,
      "flow/mag_ratio_std": 0.24318457544530364,
      "step": 6144
    },
    {
      "epoch": 0.2955983557341462,
      "grad_norm": 0.1180572658777237,
      "learning_rate": 0.00041026799522680534,
      "loss": 1.2049767971038818,
      "step": 6400
    },
    {
      "epoch": 0.30742228996351206,
      "grad_norm": 0.14954492449760437,
      "learning_rate": 0.00040285394560281816,
      "loss": 1.203136682510376,
      "step": 6656
    },
    {
      "epoch": 0.3192462241928779,
      "grad_norm": 0.27384424209594727,
      "learning_rate": 0.0003952186705398043,
      "loss": 1.2036312818527222,
      "step": 6912
    },
    {
      "epoch": 0.3310701584222438,
      "grad_norm": 0.15835484862327576,
      "learning_rate": 0.0003873732205675438,
      "loss": 1.1990163326263428,
      "step": 7168
    },
    {
      "epoch": 0.3310701584222438,
      "eval_cos_loss": 0.41148509575079567,
      "eval_loss": 1.1415519687138735,
      "eval_mse_loss": 1.1415519687138735,
      "flow/cos_sim": 0.5885149201029511,
      "flow/improvement_ratio": 0.9741524475622395,
      "flow/mag_ratio_mean": 0.5640481464122529,
      "flow/mag_ratio_std": 0.24272260563150388,
      "step": 7168
    },
    {
      "epoch": 0.3310701584222438,
      "eval_cos_loss": 0.41148509575079567,
      "eval_loss": 1.1415519687138735,
      "eval_mse_loss": 1.1415519687138735,
      "eval_runtime": 36.5211,
      "eval_samples_per_second": 766.489,
      "eval_steps_per_second": 11.993,
      "flow/cos_sim": 0.5885149201029511,
      "flow/improvement_ratio": 0.9741524475622395,
      "flow/mag_ratio_mean": 0.5640481464122529,
      "flow/mag_ratio_std": 0.24272260563150388,
      "step": 7168
    },
    {
      "epoch": 0.34289409265160964,
      "grad_norm": 0.13862715661525726,
      "learning_rate": 0.000379328950401858,
      "loss": 1.1999462842941284,
      "step": 7424
    },
    {
      "epoch": 0.3547180268809755,
      "grad_norm": 0.1664004623889923,
      "learning_rate": 0.0003710975025109345,
      "loss": 1.1982899904251099,
      "step": 7680
    },
    {
      "epoch": 0.3665419611103413,
      "grad_norm": 0.13604609668254852,
      "learning_rate": 0.0003626907902651893,
      "loss": 1.1958998441696167,
      "step": 7936
    },
    {
      "epoch": 0.37836589533970716,
      "grad_norm": 0.1457708328962326,
      "learning_rate": 0.0003541209806950514,
      "loss": 1.1945908069610596,
      "step": 8192
    },
    {
      "epoch": 0.37836589533970716,
      "eval_cos_loss": 0.4107164820851801,
      "eval_loss": 1.1382312957010312,
      "eval_mse_loss": 1.1382312957010312,
      "flow/cos_sim": 0.589283528393262,
      "flow/improvement_ratio": 0.9722015831840637,
      "flow/mag_ratio_mean": 0.5640345150477266,
      "flow/mag_ratio_std": 0.24401044291040125,
      "step": 8192
    },
    {
      "epoch": 0.37836589533970716,
      "eval_cos_loss": 0.4107164820851801,
      "eval_loss": 1.1382312957010312,
      "eval_mse_loss": 1.1382312957010312,
      "eval_runtime": 36.7013,
      "eval_samples_per_second": 762.726,
      "eval_steps_per_second": 11.934,
      "flow/cos_sim": 0.589283528393262,
      "flow/improvement_ratio": 0.9722015831840637,
      "flow/mag_ratio_mean": 0.5640345150477266,
      "flow/mag_ratio_std": 0.24401044291040125,
      "step": 8192
    },
    {
      "epoch": 0.390189829569073,
      "grad_norm": 0.16633214056491852,
      "learning_rate": 0.0003454004768816257,
      "loss": 1.1974270343780518,
      "step": 8448
    },
    {
      "epoch": 0.4020137637984389,
      "grad_norm": 0.13664524257183075,
      "learning_rate": 0.00033654190000572017,
      "loss": 1.1954563856124878,
      "step": 8704
    },
    {
      "epoch": 0.41383769802780473,
      "grad_norm": 0.17472244799137115,
      "learning_rate": 0.00032755807108121707,
      "loss": 1.1933891773223877,
      "step": 8960
    },
    {
      "epoch": 0.4256616322571706,
      "grad_norm": 0.13021035492420197,
      "learning_rate": 0.00031846199239922587,
      "loss": 1.1937364339828491,
      "step": 9216
    },
    {
      "epoch": 0.4256616322571706,
      "eval_cos_loss": 0.40818845524907654,
      "eval_loss": 1.1306110892665986,
      "eval_mse_loss": 1.1306110892665986,
      "flow/cos_sim": 0.5918115706748615,
      "flow/improvement_ratio": 0.9736669311240383,
      "flow/mag_ratio_mean": 0.5680791953382971,
      "flow/mag_ratio_std": 0.24117264903435423,
      "step": 9216
    },
    {
      "epoch": 0.4256616322571706,
      "eval_cos_loss": 0.40818845524907654,
      "eval_loss": 1.1306110892665986,
      "eval_mse_loss": 1.1306110892665986,
      "eval_runtime": 37.6562,
      "eval_samples_per_second": 743.384,
      "eval_steps_per_second": 11.632,
      "flow/cos_sim": 0.5918115706748615,
      "flow/improvement_ratio": 0.9736669311240383,
      "flow/mag_ratio_mean": 0.5680791953382971,
      "flow/mag_ratio_std": 0.24117264903435423,
      "step": 9216
    },
    {
      "epoch": 0.4374855664865364,
      "grad_norm": 0.1436239778995514,
      "learning_rate": 0.00030926682870987393,
      "loss": 1.1944515705108643,
      "step": 9472
    },
    {
      "epoch": 0.44930950071590225,
      "grad_norm": 0.1279948502779007,
      "learning_rate": 0.00029998588816897034,
      "loss": 1.192986011505127,
      "step": 9728
    },
    {
      "epoch": 0.4611334349452681,
      "grad_norm": 0.16707991063594818,
      "learning_rate": 0.00029063260307711817,
      "loss": 1.1940622329711914,
      "step": 9984
    },
    {
      "epoch": 0.472957369174634,
      "grad_norm": 0.12476647645235062,
      "learning_rate": 0.00028122051043915356,
      "loss": 1.190167784690857,
      "step": 10240
    },
    {
      "epoch": 0.472957369174634,
      "eval_cos_loss": 0.4076138965479315,
      "eval_loss": 1.1312634806110435,
      "eval_mse_loss": 1.1312634806110435,
      "flow/cos_sim": 0.5923861386296956,
      "flow/improvement_ratio": 0.9748937231764946,
      "flow/mag_ratio_mean": 0.5703548358999975,
      "flow/mag_ratio_std": 0.24083318081620622,
      "step": 10240
    },
    {
      "epoch": 0.472957369174634,
      "eval_cos_loss": 0.4076138965479315,
      "eval_loss": 1.1312634806110435,
      "eval_mse_loss": 1.1312634806110435,
      "eval_runtime": 37.3647,
      "eval_samples_per_second": 749.184,
      "eval_steps_per_second": 11.722,
      "flow/cos_sim": 0.5923861386296956,
      "flow/improvement_ratio": 0.9748937231764946,
      "flow/mag_ratio_mean": 0.5703548358999975,
      "flow/mag_ratio_std": 0.24083318081620622,
      "step": 10240
    },
    {
      "epoch": 0.48478130340399983,
      "grad_norm": 0.16375412046909332,
      "learning_rate": 0.00027176323237204403,
      "loss": 1.1899298429489136,
      "step": 10496
    },
    {
      "epoch": 0.49660523763336567,
      "grad_norm": 0.13618549704551697,
      "learning_rate": 0.0002622744563896065,
      "loss": 1.1911197900772095,
      "step": 10752
    },
    {
      "epoch": 0.5084291718627315,
      "grad_norm": 0.14637431502342224,
      "learning_rate": 0.00025276791559257494,
      "loss": 1.1925891637802124,
      "step": 11008
    },
    {
      "epoch": 0.5202531060920974,
      "grad_norm": 0.13531003892421722,
      "learning_rate": 0.0002432573687926906,
      "loss": 1.186372995376587,
      "step": 11264
    },
    {
      "epoch": 0.5202531060920974,
      "eval_cos_loss": 0.40703500609963994,
      "eval_loss": 1.1297938347407128,
      "eval_mse_loss": 1.1297938347407128,
      "flow/cos_sim": 0.5929650133603239,
      "flow/improvement_ratio": 0.9741698461308327,
      "flow/mag_ratio_mean": 0.5677392150713428,
      "flow/mag_ratio_std": 0.24396490948657468,
      "step": 11264
    },
    {
      "epoch": 0.5202531060920974,
      "eval_cos_loss": 0.40703500609963994,
      "eval_loss": 1.1297938347407128,
      "eval_mse_loss": 1.1297938347407128,
      "eval_runtime": 38.0112,
      "eval_samples_per_second": 736.441,
      "eval_steps_per_second": 11.523,
      "flow/cos_sim": 0.5929650133603239,
      "flow/improvement_ratio": 0.9741698461308327,
      "flow/mag_ratio_mean": 0.5677392150713428,
      "flow/mag_ratio_std": 0.24396490948657468,
      "step": 11264
    },
    {
      "epoch": 0.5320770403214632,
      "grad_norm": 0.13568931818008423,
      "learning_rate": 0.00023375658059958035,
      "loss": 1.1851963996887207,
      "step": 11520
    },
    {
      "epoch": 0.5439009745508291,
      "grad_norm": 0.158619686961174,
      "learning_rate": 0.00022427930149924495,
      "loss": 1.18683660030365,
      "step": 11776
    },
    {
      "epoch": 0.5557249087801949,
      "grad_norm": 0.12816773355007172,
      "learning_rate": 0.00021483924795298633,
      "loss": 1.1887577772140503,
      "step": 12032
    },
    {
      "epoch": 0.5675488430095608,
      "grad_norm": 0.1509481817483902,
      "learning_rate": 0.00020545008254558105,
      "loss": 1.1869347095489502,
      "step": 12288
    },
    {
      "epoch": 0.5675488430095608,
      "eval_cos_loss": 0.4058912583546007,
      "eval_loss": 1.1245840085695868,
      "eval_mse_loss": 1.1245840085695868,
      "flow/cos_sim": 0.5941087536888036,
      "flow/improvement_ratio": 0.9735238416009842,
      "flow/mag_ratio_mean": 0.5702283619201347,
      "flow/mag_ratio_std": 0.2440737300131419,
      "step": 12288
    },
    {
      "epoch": 0.5675488430095608,
      "eval_cos_loss": 0.4058912583546007,
      "eval_loss": 1.1245840085695868,
      "eval_mse_loss": 1.1245840085695868,
      "eval_runtime": 37.443,
      "eval_samples_per_second": 747.617,
      "eval_steps_per_second": 11.698,
      "flow/cos_sim": 0.5941087536888036,
      "flow/improvement_ratio": 0.9735238416009842,
      "flow/mag_ratio_mean": 0.5702283619201347,
      "flow/mag_ratio_std": 0.2440737300131419,
      "step": 12288
    },
    {
      "epoch": 0.5793727772389267,
      "grad_norm": 0.12331920862197876,
      "learning_rate": 0.00019612539421142756,
      "loss": 1.1875509023666382,
      "step": 12544
    },
    {
      "epoch": 0.5911967114682924,
      "grad_norm": 0.14404380321502686,
      "learning_rate": 0.00018687867856728862,
      "loss": 1.186294436454773,
      "step": 12800
    },
    {
      "epoch": 0.6030206456976583,
      "grad_norm": 0.16519756615161896,
      "learning_rate": 0.00017772331838009138,
      "loss": 1.1816554069519043,
      "step": 13056
    },
    {
      "epoch": 0.6148445799270241,
      "grad_norm": 0.1261790543794632,
      "learning_rate": 0.00016867256419805628,
      "loss": 1.184910774230957,
      "step": 13312
    },
    {
      "epoch": 0.6148445799270241,
      "eval_cos_loss": 0.4058965126129046,
      "eval_loss": 1.124463065301991,
      "eval_mse_loss": 1.124463065301991,
      "flow/cos_sim": 0.5941034868427607,
      "flow/improvement_ratio": 0.9736478256580492,
      "flow/mag_ratio_mean": 0.5693308626680069,
      "flow/mag_ratio_std": 0.24294855394592024,
      "step": 13312
    },
    {
      "epoch": 0.6148445799270241,
      "eval_cos_loss": 0.4058965126129046,
      "eval_loss": 1.124463065301991,
      "eval_mse_loss": 1.124463065301991,
      "eval_runtime": 37.6078,
      "eval_samples_per_second": 744.34,
      "eval_steps_per_second": 11.647,
      "flow/cos_sim": 0.5941034868427607,
      "flow/improvement_ratio": 0.9736478256580492,
      "flow/mag_ratio_mean": 0.5693308626680069,
      "flow/mag_ratio_std": 0.24294855394592024,
      "step": 13312
    },
    {
      "epoch": 0.62666851415639,
      "grad_norm": 0.10855361819267273,
      "learning_rate": 0.00015973951517318435,
      "loss": 1.1825791597366333,
      "step": 13568
    },
    {
      "epoch": 0.6384924483857558,
      "grad_norm": 0.12764589488506317,
      "learning_rate": 0.00015093710010286203,
      "loss": 1.183720588684082,
      "step": 13824
    },
    {
      "epoch": 0.6503163826151217,
      "grad_norm": 0.15194828808307648,
      "learning_rate": 0.00014227805871801813,
      "loss": 1.1843342781066895,
      "step": 14080
    },
    {
      "epoch": 0.6621403168444876,
      "grad_norm": 0.1313941776752472,
      "learning_rate": 0.00013377492324491864,
      "loss": 1.1845036745071411,
      "step": 14336
    },
    {
      "epoch": 0.6621403168444876,
      "eval_cos_loss": 0.4055524829043645,
      "eval_loss": 1.1236549692066837,
      "eval_mse_loss": 1.1236549692066837,
      "flow/cos_sim": 0.5944475390051054,
      "flow/improvement_ratio": 0.9741561509430681,
      "flow/mag_ratio_mean": 0.5722755660477294,
      "flow/mag_ratio_std": 0.24292387908588262,
      "step": 14336
    },
    {
      "epoch": 0.6621403168444876,
      "eval_cos_loss": 0.4055524829043645,
      "eval_loss": 1.1236549692066837,
      "eval_mse_loss": 1.1236549692066837,
      "eval_runtime": 37.766,
      "eval_samples_per_second": 741.222,
      "eval_steps_per_second": 11.598,
      "flow/cos_sim": 0.5944475390051054,
      "flow/improvement_ratio": 0.9741561509430681,
      "flow/mag_ratio_mean": 0.5722755660477294,
      "flow/mag_ratio_std": 0.24292387908588262,
      "step": 14336
    },
    {
      "epoch": 0.6739642510738534,
      "grad_norm": 0.1419185996055603,
      "learning_rate": 0.00012544000026728114,
      "loss": 1.1815910339355469,
      "step": 14592
    },
    {
      "epoch": 0.6857881853032193,
      "grad_norm": 0.11921229213476181,
      "learning_rate": 0.00011728535291496281,
      "loss": 1.1859562397003174,
      "step": 14848
    },
    {
      "epoch": 0.6976121195325851,
      "grad_norm": 0.13644331693649292,
      "learning_rate": 0.00010932278340499846,
      "loss": 1.1809968948364258,
      "step": 15104
    },
    {
      "epoch": 0.709436053761951,
      "grad_norm": 0.12353217601776123,
      "learning_rate": 0.0001015638159602576,
      "loss": 1.1835533380508423,
      "step": 15360
    },
    {
      "epoch": 0.709436053761951,
      "eval_cos_loss": 0.404491749596378,
      "eval_loss": 1.1212830510857987,
      "eval_mse_loss": 1.1212830510857987,
      "flow/cos_sim": 0.5955082661212852,
      "flow/improvement_ratio": 0.9744905962791617,
      "flow/mag_ratio_mean": 0.5719217330081278,
      "flow/mag_ratio_std": 0.24169711207282052,
      "step": 15360
    },
    {
      "epoch": 0.709436053761951,
      "eval_cos_loss": 0.404491749596378,
      "eval_loss": 1.1212830510857987,
      "eval_mse_loss": 1.1212830510857987,
      "eval_runtime": 37.817,
      "eval_samples_per_second": 740.223,
      "eval_steps_per_second": 11.582,
      "flow/cos_sim": 0.5955082661212852,
      "flow/improvement_ratio": 0.9744905962791617,
      "flow/mag_ratio_mean": 0.5719217330081278,
      "flow/mag_ratio_std": 0.24169711207282052,
      "step": 15360
    },
    {
      "epoch": 0.7212599879913169,
      "grad_norm": 0.11799845844507217,
      "learning_rate": 9.401968013044271e-05,
      "loss": 1.1806869506835938,
      "step": 15616
    },
    {
      "epoch": 0.7330839222206826,
      "grad_norm": 0.12838932871818542,
      "learning_rate": 8.670129453956732e-05,
      "loss": 1.182599663734436,
      "step": 15872
    },
    {
      "epoch": 0.7449078564500485,
      "grad_norm": 0.11126351356506348,
      "learning_rate": 7.961925108343715e-05,
      "loss": 1.1809351444244385,
      "step": 16128
    },
    {
      "epoch": 0.7567317906794143,
      "grad_norm": 0.13469421863555908,
      "learning_rate": 7.278379960000436e-05,
      "loss": 1.1837177276611328,
      "step": 16384
    },
    {
      "epoch": 0.7567317906794143,
      "eval_cos_loss": 0.40498853165265086,
      "eval_loss": 1.1224167082952037,
      "eval_mse_loss": 1.1224167082952037,
      "flow/cos_sim": 0.5950114819557155,
      "flow/improvement_ratio": 0.9733862189669588,
      "flow/mag_ratio_mean": 0.5712394598684355,
      "flow/mag_ratio_std": 0.24294140764839572,
      "step": 16384
    },
    {
      "epoch": 0.7567317906794143,
      "eval_cos_loss": 0.40498853165265086,
      "eval_loss": 1.1224167082952037,
      "eval_mse_loss": 1.1224167082952037,
      "eval_runtime": 38.0074,
      "eval_samples_per_second": 736.514,
      "eval_steps_per_second": 11.524,
      "flow/cos_sim": 0.5950114819557155,
      "flow/improvement_ratio": 0.9733862189669588,
      "flow/mag_ratio_mean": 0.5712394598684355,
      "flow/mag_ratio_std": 0.24294140764839572,
      "step": 16384
    },
    {
      "epoch": 0.7685557249087802,
      "grad_norm": 0.13311493396759033,
      "learning_rate": 6.62048330347825e-05,
      "loss": 1.1809964179992676,
      "step": 16640
    },
    {
      "epoch": 0.780379659138146,
      "grad_norm": 0.10799466073513031,
      "learning_rate": 5.989187312279115e-05,
      "loss": 1.180505633354187,
      "step": 16896
    },
    {
      "epoch": 0.7922035933675119,
      "grad_norm": 0.1486814022064209,
      "learning_rate": 5.3854056607753746e-05,
      "loss": 1.1795099973678589,
      "step": 17152
    },
    {
      "epoch": 0.8040275275968778,
      "grad_norm": 0.11168931424617767,
      "learning_rate": 4.8100122018492956e-05,
      "loss": 1.1790964603424072,
      "step": 17408
    },
    {
      "epoch": 0.8040275275968778,
      "eval_cos_loss": 0.40403676060236754,
      "eval_loss": 1.1197874216728558,
      "eval_mse_loss": 1.1197874216728558,
      "flow/cos_sim": 0.5959632730102975,
      "flow/improvement_ratio": 0.974381895370135,
      "flow/mag_ratio_mean": 0.5709017501574128,
      "flow/mag_ratio_std": 0.24267656927647657,
      "step": 17408
    },
    {
      "epoch": 0.8040275275968778,
      "eval_cos_loss": 0.40403676060236754,
      "eval_loss": 1.1197874216728558,
      "eval_mse_loss": 1.1197874216728558,
      "eval_runtime": 37.0697,
      "eval_samples_per_second": 755.145,
      "eval_steps_per_second": 11.816,
      "flow/cos_sim": 0.5959632730102975,
      "flow/improvement_ratio": 0.974381895370135,
      "flow/mag_ratio_mean": 0.5709017501574128,
      "flow/mag_ratio_std": 0.24267656927647657,
      "step": 17408
    },
    {
      "epoch": 0.8158514618262436,
      "grad_norm": 0.120413638651371,
      "learning_rate": 4.263839702166214e-05,
      "loss": 1.1833932399749756,
      "step": 17664
    },
    {
      "epoch": 0.8276753960556095,
      "grad_norm": 0.13419309258460999,
      "learning_rate": 3.747678636911772e-05,
      "loss": 1.1790565252304077,
      "step": 17920
    },
    {
      "epoch": 0.8394993302849753,
      "grad_norm": 0.13175411522388458,
      "learning_rate": 3.262276045737592e-05,
      "loss": 1.182090401649475,
      "step": 18176
    },
    {
      "epoch": 0.8513232645143411,
      "grad_norm": 0.09947279095649719,
      "learning_rate": 2.8083344515711012e-05,
      "loss": 1.177965760231018,
      "step": 18432
    },
    {
      "epoch": 0.8513232645143411,
      "eval_cos_loss": 0.40370703621269904,
      "eval_loss": 1.1215725149193856,
      "eval_mse_loss": 1.1215725149193856,
      "flow/cos_sim": 0.5962929766472072,
      "flow/improvement_ratio": 0.9753795086520992,
      "flow/mag_ratio_mean": 0.5721857149034875,
      "flow/mag_ratio_std": 0.24184994011709135,
      "step": 18432
    },
    {
      "epoch": 0.8513232645143411,
      "eval_cos_loss": 0.40370703621269904,
      "eval_loss": 1.1215725149193856,
      "eval_mse_loss": 1.1215725149193856,
      "eval_runtime": 37.003,
      "eval_samples_per_second": 756.506,
      "eval_steps_per_second": 11.837,
      "flow/cos_sim": 0.5962929766472072,
      "flow/improvement_ratio": 0.9753795086520992,
      "flow/mag_ratio_mean": 0.5721857149034875,
      "flow/mag_ratio_std": 0.24184994011709135,
      "step": 18432
    },
    {
      "epoch": 0.8631471987437069,
      "grad_norm": 0.11105956137180328,
      "learning_rate": 2.3865108438545337e-05,
      "loss": 1.181681752204895,
      "step": 18688
    },
    {
      "epoch": 0.8749711329730728,
      "grad_norm": 0.10993946343660355,
      "learning_rate": 1.9974157276843596e-05,
      "loss": 1.1794625520706177,
      "step": 18944
    },
    {
      "epoch": 0.8867950672024387,
      "grad_norm": 0.10621558874845505,
      "learning_rate": 1.641612240227641e-05,
      "loss": 1.1792101860046387,
      "step": 19200
    },
    {
      "epoch": 0.8986190014318045,
      "grad_norm": 0.09712148457765579,
      "learning_rate": 1.3196153356938134e-05,
      "loss": 1.1793529987335205,
      "step": 19456
    },
    {
      "epoch": 0.8986190014318045,
      "eval_cos_loss": 0.40458166646902965,
      "eval_loss": 1.1218491670203536,
      "eval_mse_loss": 1.1218491670203536,
      "flow/cos_sim": 0.5954183448939563,
      "flow/improvement_ratio": 0.9741832541820665,
      "flow/mag_ratio_mean": 0.5714860147537162,
      "flow/mag_ratio_std": 0.2427709357787485,
      "step": 19456
    },
    {
      "epoch": 0.8986190014318045,
      "eval_cos_loss": 0.40458166646902965,
      "eval_loss": 1.1218491670203536,
      "eval_mse_loss": 1.1218491670203536,
      "eval_runtime": 37.3338,
      "eval_samples_per_second": 749.804,
      "eval_steps_per_second": 11.732,
      "flow/cos_sim": 0.5954183448939563,
      "flow/improvement_ratio": 0.9741832541820665,
      "flow/mag_ratio_mean": 0.5714860147537162,
      "flow/mag_ratio_std": 0.2427709357787485,
      "step": 19456
    },
    {
      "epoch": 0.9104429356611704,
      "grad_norm": 0.12275788933038712,
      "learning_rate": 1.031891040041788e-05,
      "loss": 1.1787521839141846,
      "step": 19712
    },
    {
      "epoch": 0.9222668698905362,
      "grad_norm": 0.1388554871082306,
      "learning_rate": 7.788557765007614e-06,
      "loss": 1.1809200048446655,
      "step": 19968
    },
    {
      "epoch": 0.9340908041199021,
      "grad_norm": 0.11414396017789841,
      "learning_rate": 5.6087576288111344e-06,
      "loss": 1.1808236837387085,
      "step": 20224
    },
    {
      "epoch": 0.945914738349268,
      "grad_norm": 0.11537494510412216,
      "learning_rate": 3.782664815475373e-06,
      "loss": 1.1786174774169922,
      "step": 20480
    },
    {
      "epoch": 0.945914738349268,
      "eval_cos_loss": 0.4043068891246569,
      "eval_loss": 1.120300586909464,
      "eval_mse_loss": 1.120300586909464,
      "flow/cos_sim": 0.595693129110554,
      "flow/improvement_ratio": 0.9729666603754644,
      "flow/mag_ratio_mean": 0.5716023456016087,
      "flow/mag_ratio_std": 0.24340959124641331,
      "step": 20480
    },
    {
      "epoch": 0.945914738349268,
      "eval_cos_loss": 0.4043068891246569,
      "eval_loss": 1.120300586909464,
      "eval_mse_loss": 1.120300586909464,
      "eval_runtime": 37.0085,
      "eval_samples_per_second": 756.394,
      "eval_steps_per_second": 11.835,
      "flow/cos_sim": 0.595693129110554,
      "flow/improvement_ratio": 0.9729666603754644,
      "flow/mag_ratio_mean": 0.5716023456016087,
      "flow/mag_ratio_std": 0.24340959124641331,
      "step": 20480
    },
    {
      "epoch": 0.9577386725786338,
      "grad_norm": 0.12133973091840744,
      "learning_rate": 2.31292222821583e-06,
      "loss": 1.1803685426712036,
      "step": 20736
    },
    {
      "epoch": 0.9695626068079997,
      "grad_norm": 0.10521073639392853,
      "learning_rate": 1.201657024744035e-06,
      "loss": 1.1818820238113403,
      "step": 20992
    },
    {
      "epoch": 0.9813865410373654,
      "grad_norm": 0.09989389777183533,
      "learning_rate": 4.5047753863319827e-07,
      "loss": 1.1803994178771973,
      "step": 21248
    },
    {
      "epoch": 0.9932104752667313,
      "grad_norm": 0.09399425238370895,
      "learning_rate": 6.047095157787896e-08,
      "loss": 1.1804349422454834,
      "step": 21504
    },
    {
      "epoch": 0.9932104752667313,
      "eval_cos_loss": 0.40440767044073916,
      "eval_loss": 1.1205239834850782,
      "eval_mse_loss": 1.1205239834850782,
      "flow/cos_sim": 0.5955923152024343,
      "flow/improvement_ratio": 0.9742439836944075,
      "flow/mag_ratio_mean": 0.5713475014521107,
      "flow/mag_ratio_std": 0.24330289396521163,
      "step": 21504
    },
    {
      "epoch": 0.9932104752667313,
      "eval_cos_loss": 0.40440767044073916,
      "eval_loss": 1.1205239834850782,
      "eval_mse_loss": 1.1205239834850782,
      "eval_runtime": 37.1012,
      "eval_samples_per_second": 754.504,
      "eval_steps_per_second": 11.806,
      "flow/cos_sim": 0.5955923152024343,
      "flow/improvement_ratio": 0.9742439836944075,
      "flow/mag_ratio_mean": 0.5713475014521107,
      "flow/mag_ratio_std": 0.24330289396521163,
      "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
}