{
  "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": 0.3170062303543091,
      "learning_rate": 0.000498046875,
      "loss": 2.42356538772583,
      "step": 256
    },
    {
      "epoch": 0.023647868458731697,
      "grad_norm": 0.20766793191432953,
      "learning_rate": 0.000998046875,
      "loss": 2.0067272186279297,
      "step": 512
    },
    {
      "epoch": 0.03547180268809755,
      "grad_norm": 0.47662925720214844,
      "learning_rate": 0.000999640996023194,
      "loss": 1.8047175407409668,
      "step": 768
    },
    {
      "epoch": 0.047295736917463395,
      "grad_norm": 0.38411229848861694,
      "learning_rate": 0.0009985588674043958,
      "loss": 1.7324724197387695,
      "step": 1024
    },
    {
      "epoch": 0.047295736917463395,
      "eval_cos_loss": 0.4818032477136072,
      "eval_loss": 1.6560364654075066,
      "eval_mse_loss": 1.4151348444424807,
      "flow/cos_sim": 0.5181967507894725,
      "flow/improvement_ratio": 0.954772694197964,
      "flow/mag_ratio_mean": 0.48862707193039323,
      "flow/mag_ratio_std": 0.18404143276416005,
      "step": 1024
    },
    {
      "epoch": 0.047295736917463395,
      "eval_cos_loss": 0.4818032477136072,
      "eval_loss": 1.6560364654075066,
      "eval_mse_loss": 1.4151348444424807,
      "eval_runtime": 41.6432,
      "eval_samples_per_second": 672.211,
      "eval_steps_per_second": 10.518,
      "flow/cos_sim": 0.5181967507894725,
      "flow/improvement_ratio": 0.954772694197964,
      "flow/mag_ratio_mean": 0.48862707193039323,
      "flow/mag_ratio_std": 0.18404143276416005,
      "step": 1024
    },
    {
      "epoch": 0.05911967114682925,
      "grad_norm": 0.4044971168041229,
      "learning_rate": 0.0009967551747861387,
      "loss": 1.6944022178649902,
      "step": 1280
    },
    {
      "epoch": 0.0709436053761951,
      "grad_norm": 0.23417921364307404,
      "learning_rate": 0.000994232528651847,
      "loss": 1.6682453155517578,
      "step": 1536
    },
    {
      "epoch": 0.08276753960556095,
      "grad_norm": 0.11558878421783447,
      "learning_rate": 0.0009909945800260092,
      "loss": 1.6490408182144165,
      "step": 1792
    },
    {
      "epoch": 0.09459147383492679,
      "grad_norm": 0.1645389050245285,
      "learning_rate": 0.0009870460151900522,
      "loss": 1.6396121978759766,
      "step": 2048
    },
    {
      "epoch": 0.09459147383492679,
      "eval_cos_loss": 0.4538577257498214,
      "eval_loss": 1.5759433782808312,
      "eval_mse_loss": 1.3490145179234683,
      "flow/cos_sim": 0.5461422990174054,
      "flow/improvement_ratio": 0.9588973424206041,
      "flow/mag_ratio_mean": 0.5239204951481188,
      "flow/mag_ratio_std": 0.19641730230148524,
      "step": 2048
    },
    {
      "epoch": 0.09459147383492679,
      "eval_cos_loss": 0.4538577257498214,
      "eval_loss": 1.5759433782808312,
      "eval_mse_loss": 1.3490145179234683,
      "eval_runtime": 38.1827,
      "eval_samples_per_second": 733.134,
      "eval_steps_per_second": 11.471,
      "flow/cos_sim": 0.5461422990174054,
      "flow/improvement_ratio": 0.9588973424206041,
      "flow/mag_ratio_mean": 0.5239204951481188,
      "flow/mag_ratio_std": 0.19641730230148524,
      "step": 2048
    },
    {
      "epoch": 0.10641540806429264,
      "grad_norm": 0.1483554244041443,
      "learning_rate": 0.0009823925488998885,
      "loss": 1.629364252090454,
      "step": 2304
    },
    {
      "epoch": 0.1182393422936585,
      "grad_norm": 0.1272267997264862,
      "learning_rate": 0.0009770409161149525,
      "loss": 1.6207596063613892,
      "step": 2560
    },
    {
      "epoch": 0.13006327652302435,
      "grad_norm": 0.4405772387981415,
      "learning_rate": 0.0009709988622506973,
      "loss": 1.6130421161651611,
      "step": 2816
    },
    {
      "epoch": 0.1418872107523902,
      "grad_norm": 0.2765417993068695,
      "learning_rate": 0.000964275131968659,
      "loss": 1.6063523292541504,
      "step": 3072
    },
    {
      "epoch": 0.1418872107523902,
      "eval_cos_loss": 0.4437784091930955,
      "eval_loss": 1.5468863191669935,
      "eval_mse_loss": 1.3249971137199228,
      "flow/cos_sim": 0.5562216001286354,
      "flow/improvement_ratio": 0.9605493699306766,
      "flow/mag_ratio_mean": 0.531293115863517,
      "flow/mag_ratio_std": 0.20344179882306487,
      "step": 3072
    },
    {
      "epoch": 0.1418872107523902,
      "eval_cos_loss": 0.4437784091930955,
      "eval_loss": 1.5468863191669935,
      "eval_mse_loss": 1.3249971137199228,
      "eval_runtime": 38.9587,
      "eval_samples_per_second": 718.531,
      "eval_steps_per_second": 11.243,
      "flow/cos_sim": 0.5562216001286354,
      "flow/improvement_ratio": 0.9605493699306766,
      "flow/mag_ratio_mean": 0.531293115863517,
      "flow/mag_ratio_std": 0.20344179882306487,
      "step": 3072
    },
    {
      "epoch": 0.15371114498175603,
      "grad_norm": 0.14539985358715057,
      "learning_rate": 0.0009568794565203123,
      "loss": 1.6010686159133911,
      "step": 3328
    },
    {
      "epoch": 0.1655350792111219,
      "grad_norm": 0.146505206823349,
      "learning_rate": 0.0009488225396630347,
      "loss": 1.593965768814087,
      "step": 3584
    },
    {
      "epoch": 0.17735901344048774,
      "grad_norm": 0.17704616487026215,
      "learning_rate": 0.0009401160421685646,
      "loss": 1.591158390045166,
      "step": 3840
    },
    {
      "epoch": 0.18918294766985358,
      "grad_norm": 0.21900475025177002,
      "learning_rate": 0.0009307725649463714,
      "loss": 1.5863914489746094,
      "step": 4096
    },
    {
      "epoch": 0.18918294766985358,
      "eval_cos_loss": 0.4372148915240754,
      "eval_loss": 1.5282805349728832,
      "eval_mse_loss": 1.3096730913201424,
      "flow/cos_sim": 0.5627851008552395,
      "flow/improvement_ratio": 0.960628461483951,
      "flow/mag_ratio_mean": 0.5383912115336553,
      "flow/mag_ratio_std": 0.20601355281050346,
      "step": 4096
    },
    {
      "epoch": 0.18918294766985358,
      "eval_cos_loss": 0.4372148915240754,
      "eval_loss": 1.5282805349728832,
      "eval_mse_loss": 1.3096730913201424,
      "eval_runtime": 37.5818,
      "eval_samples_per_second": 744.854,
      "eval_steps_per_second": 11.655,
      "flow/cos_sim": 0.5627851008552395,
      "flow/improvement_ratio": 0.960628461483951,
      "flow/mag_ratio_mean": 0.5383912115336553,
      "flow/mag_ratio_std": 0.20601355281050346,
      "step": 4096
    },
    {
      "epoch": 0.20100688189921945,
      "grad_norm": 0.3215397000312805,
      "learning_rate": 0.0009208056308063659,
      "loss": 1.5831135511398315,
      "step": 4352
    },
    {
      "epoch": 0.2128308161285853,
      "grad_norm": 0.39093902707099915,
      "learning_rate": 0.0009102296648873445,
      "loss": 1.5793178081512451,
      "step": 4608
    },
    {
      "epoch": 0.22465475035795113,
      "grad_norm": 0.24414080381393433,
      "learning_rate": 0.0008990599737794927,
      "loss": 1.5755181312561035,
      "step": 4864
    },
    {
      "epoch": 0.236478684587317,
      "grad_norm": 0.18973594903945923,
      "learning_rate": 0.0008873127233711644,
      "loss": 1.5741870403289795,
      "step": 5120
    },
    {
      "epoch": 0.236478684587317,
      "eval_cos_loss": 0.43135033779220494,
      "eval_loss": 1.5102224085973277,
      "eval_mse_loss": 1.294547238578535,
      "flow/cos_sim": 0.568649673570781,
      "flow/improvement_ratio": 0.9623238818558384,
      "flow/mag_ratio_mean": 0.5475505786399318,
      "flow/mag_ratio_std": 0.2083817292089876,
      "step": 5120
    },
    {
      "epoch": 0.236478684587317,
      "eval_cos_loss": 0.43135033779220494,
      "eval_loss": 1.5102224085973277,
      "eval_mse_loss": 1.294547238578535,
      "eval_runtime": 38.3814,
      "eval_samples_per_second": 729.337,
      "eval_steps_per_second": 11.412,
      "flow/cos_sim": 0.568649673570781,
      "flow/improvement_ratio": 0.9623238818558384,
      "flow/mag_ratio_mean": 0.5475505786399318,
      "flow/mag_ratio_std": 0.2083817292089876,
      "step": 5120
    },
    {
      "epoch": 0.24830261881668284,
      "grad_norm": 0.23730266094207764,
      "learning_rate": 0.0008750049154520011,
      "loss": 1.5687248706817627,
      "step": 5376
    },
    {
      "epoch": 0.2601265530460487,
      "grad_norm": 0.23536692559719086,
      "learning_rate": 0.0008621543631062487,
      "loss": 1.5671372413635254,
      "step": 5632
    },
    {
      "epoch": 0.27195048727541454,
      "grad_norm": 0.15165014564990997,
      "learning_rate": 0.0008487796649318904,
      "loss": 1.5671967267990112,
      "step": 5888
    },
    {
      "epoch": 0.2837744215047804,
      "grad_norm": 0.1857716292142868,
      "learning_rate": 0.0008349001781229053,
      "loss": 1.5656864643096924,
      "step": 6144
    },
    {
      "epoch": 0.2837744215047804,
      "eval_cos_loss": 0.4303430036458795,
      "eval_loss": 1.507517477148744,
      "eval_mse_loss": 1.292345978353666,
      "flow/cos_sim": 0.5696570243193134,
      "flow/improvement_ratio": 0.9619137474663182,
      "flow/mag_ratio_mean": 0.5517072507507725,
      "flow/mag_ratio_std": 0.2047896740711443,
      "step": 6144
    },
    {
      "epoch": 0.2837744215047804,
      "eval_cos_loss": 0.4303430036458795,
      "eval_loss": 1.507517477148744,
      "eval_mse_loss": 1.292345978353666,
      "eval_runtime": 39.1056,
      "eval_samples_per_second": 715.83,
      "eval_steps_per_second": 11.2,
      "flow/cos_sim": 0.5696570243193134,
      "flow/improvement_ratio": 0.9619137474663182,
      "flow/mag_ratio_mean": 0.5517072507507725,
      "flow/mag_ratio_std": 0.2047896740711443,
      "step": 6144
    },
    {
      "epoch": 0.2955983557341462,
      "grad_norm": 0.16939695179462433,
      "learning_rate": 0.0008205359904536107,
      "loss": 1.5614818334579468,
      "step": 6400
    },
    {
      "epoch": 0.30742228996351206,
      "grad_norm": 0.20279403030872345,
      "learning_rate": 0.0008057078912056363,
      "loss": 1.5598734617233276,
      "step": 6656
    },
    {
      "epoch": 0.3192462241928779,
      "grad_norm": 0.17468483746051788,
      "learning_rate": 0.0007904373410796086,
      "loss": 1.5608937740325928,
      "step": 6912
    },
    {
      "epoch": 0.3310701584222438,
      "grad_norm": 0.13540241122245789,
      "learning_rate": 0.0007747464411350876,
      "loss": 1.5515446662902832,
      "step": 7168
    },
    {
      "epoch": 0.3310701584222438,
      "eval_cos_loss": 0.42760413388411206,
      "eval_loss": 1.4991425577908346,
      "eval_mse_loss": 1.285340489862172,
      "flow/cos_sim": 0.5723958823778857,
      "flow/improvement_ratio": 0.9629914771476292,
      "flow/mag_ratio_mean": 0.551068566841622,
      "flow/mag_ratio_std": 0.21019490124428108,
      "step": 7168
    },
    {
      "epoch": 0.3310701584222438,
      "eval_cos_loss": 0.42760413388411206,
      "eval_loss": 1.4991425577908346,
      "eval_mse_loss": 1.285340489862172,
      "eval_runtime": 37.5914,
      "eval_samples_per_second": 744.665,
      "eval_steps_per_second": 11.652,
      "flow/cos_sim": 0.5723958823778857,
      "flow/improvement_ratio": 0.9629914771476292,
      "flow/mag_ratio_mean": 0.551068566841622,
      "flow/mag_ratio_std": 0.21019490124428108,
      "step": 7168
    },
    {
      "epoch": 0.34289409265160964,
      "grad_norm": 0.16203606128692627,
      "learning_rate": 0.000758657900803716,
      "loss": 1.5541112422943115,
      "step": 7424
    },
    {
      "epoch": 0.3547180268809755,
      "grad_norm": 0.15600702166557312,
      "learning_rate": 0.000742195005021869,
      "loss": 1.552086353302002,
      "step": 7680
    },
    {
      "epoch": 0.3665419611103413,
      "grad_norm": 0.25274714827537537,
      "learning_rate": 0.0007253815805303786,
      "loss": 1.5514739751815796,
      "step": 7936
    },
    {
      "epoch": 0.37836589533970716,
      "grad_norm": 0.16391786932945251,
      "learning_rate": 0.0007082419613901028,
      "loss": 1.5471858978271484,
      "step": 8192
    },
    {
      "epoch": 0.37836589533970716,
      "eval_cos_loss": 0.42565103276679506,
      "eval_loss": 1.4905888979837774,
      "eval_mse_loss": 1.277763385206597,
      "flow/cos_sim": 0.5743489830189099,
      "flow/improvement_ratio": 0.9618241816895193,
      "flow/mag_ratio_mean": 0.5533584926498535,
      "flow/mag_ratio_std": 0.21572684449011878,
      "step": 8192
    },
    {
      "epoch": 0.37836589533970716,
      "eval_cos_loss": 0.42565103276679506,
      "eval_loss": 1.4905888979837774,
      "eval_mse_loss": 1.277763385206597,
      "eval_runtime": 37.3487,
      "eval_samples_per_second": 749.503,
      "eval_steps_per_second": 11.727,
      "flow/cos_sim": 0.5743489830189099,
      "flow/improvement_ratio": 0.9618241816895193,
      "flow/mag_ratio_mean": 0.5533584926498535,
      "flow/mag_ratio_std": 0.21572684449011878,
      "step": 8192
    },
    {
      "epoch": 0.390189829569073,
      "grad_norm": 0.13766784965991974,
      "learning_rate": 0.0006908009537632514,
      "loss": 1.5495105981826782,
      "step": 8448
    },
    {
      "epoch": 0.4020137637984389,
      "grad_norm": 0.15862250328063965,
      "learning_rate": 0.0006730838000114403,
      "loss": 1.5457631349563599,
      "step": 8704
    },
    {
      "epoch": 0.41383769802780473,
      "grad_norm": 0.1839229315519333,
      "learning_rate": 0.0006551161421624341,
      "loss": 1.5441044569015503,
      "step": 8960
    },
    {
      "epoch": 0.4256616322571706,
      "grad_norm": 0.13566677272319794,
      "learning_rate": 0.0006369239847984517,
      "loss": 1.5432497262954712,
      "step": 9216
    },
    {
      "epoch": 0.4256616322571706,
      "eval_cos_loss": 0.42430811161047793,
      "eval_loss": 1.4869584389473205,
      "eval_mse_loss": 1.2748043828358933,
      "flow/cos_sim": 0.5756919168990496,
      "flow/improvement_ratio": 0.96208056807518,
      "flow/mag_ratio_mean": 0.5574794498752785,
      "flow/mag_ratio_std": 0.21047259782139024,
      "step": 9216
    },
    {
      "epoch": 0.4256616322571706,
      "eval_cos_loss": 0.42430811161047793,
      "eval_loss": 1.4869584389473205,
      "eval_mse_loss": 1.2748043828358933,
      "eval_runtime": 37.5205,
      "eval_samples_per_second": 746.072,
      "eval_steps_per_second": 11.674,
      "flow/cos_sim": 0.5756919168990496,
      "flow/improvement_ratio": 0.96208056807518,
      "flow/mag_ratio_mean": 0.5574794498752785,
      "flow/mag_ratio_std": 0.21047259782139024,
      "step": 9216
    },
    {
      "epoch": 0.4374855664865364,
      "grad_norm": 0.13037081062793732,
      "learning_rate": 0.0006185336574197479,
      "loss": 1.544568419456482,
      "step": 9472
    },
    {
      "epoch": 0.44930950071590225,
      "grad_norm": 0.2653915286064148,
      "learning_rate": 0.0005999717763379407,
      "loss": 1.5420877933502197,
      "step": 9728
    },
    {
      "epoch": 0.4611334349452681,
      "grad_norm": 0.12325401604175568,
      "learning_rate": 0.0005812652061542363,
      "loss": 1.5435826778411865,
      "step": 9984
    },
    {
      "epoch": 0.472957369174634,
      "grad_norm": 0.2154727578163147,
      "learning_rate": 0.0005624410208783071,
      "loss": 1.5385282039642334,
      "step": 10240
    },
    {
      "epoch": 0.472957369174634,
      "eval_cos_loss": 0.42216237005033447,
      "eval_loss": 1.4820257316441297,
      "eval_mse_loss": 1.2709445436251219,
      "flow/cos_sim": 0.577837641925028,
      "flow/improvement_ratio": 0.9637134710675506,
      "flow/mag_ratio_mean": 0.5554149469284162,
      "flow/mag_ratio_std": 0.2115786727432791,
      "step": 10240
    },
    {
      "epoch": 0.472957369174634,
      "eval_cos_loss": 0.42216237005033447,
      "eval_loss": 1.4820257316441297,
      "eval_mse_loss": 1.2709445436251219,
      "eval_runtime": 37.3144,
      "eval_samples_per_second": 750.194,
      "eval_steps_per_second": 11.738,
      "flow/cos_sim": 0.577837641925028,
      "flow/improvement_ratio": 0.9637134710675506,
      "flow/mag_ratio_mean": 0.5554149469284162,
      "flow/mag_ratio_std": 0.2115786727432791,
      "step": 10240
    },
    {
      "epoch": 0.48478130340399983,
      "grad_norm": 0.167108952999115,
      "learning_rate": 0.0005435264647440881,
      "loss": 1.5393356084823608,
      "step": 10496
    },
    {
      "epoch": 0.49660523763336567,
      "grad_norm": 0.11320369690656662,
      "learning_rate": 0.000524548912779213,
      "loss": 1.5383528470993042,
      "step": 10752
    },
    {
      "epoch": 0.5084291718627315,
      "grad_norm": 0.279090940952301,
      "learning_rate": 0.0005055358311851499,
      "loss": 1.5419063568115234,
      "step": 11008
    },
    {
      "epoch": 0.5202531060920974,
      "grad_norm": 0.1590709239244461,
      "learning_rate": 0.0004865147375853812,
      "loss": 1.5351711511611938,
      "step": 11264
    },
    {
      "epoch": 0.5202531060920974,
      "eval_cos_loss": 0.42213277620811984,
      "eval_loss": 1.4814584206228387,
      "eval_mse_loss": 1.2703920295249382,
      "flow/cos_sim": 0.577867228690892,
      "flow/improvement_ratio": 0.962785639174997,
      "flow/mag_ratio_mean": 0.5578808011529652,
      "flow/mag_ratio_std": 0.21099132279010668,
      "step": 11264
    },
    {
      "epoch": 0.5202531060920974,
      "eval_cos_loss": 0.42213277620811984,
      "eval_loss": 1.4814584206228387,
      "eval_mse_loss": 1.2703920295249382,
      "eval_runtime": 38.5165,
      "eval_samples_per_second": 726.78,
      "eval_steps_per_second": 11.372,
      "flow/cos_sim": 0.577867228690892,
      "flow/improvement_ratio": 0.962785639174997,
      "flow/mag_ratio_mean": 0.5578808011529652,
      "flow/mag_ratio_std": 0.21099132279010668,
      "step": 11264
    },
    {
      "epoch": 0.5320770403214632,
      "grad_norm": 0.14667300879955292,
      "learning_rate": 0.0004675131611991607,
      "loss": 1.5333118438720703,
      "step": 11520
    },
    {
      "epoch": 0.5439009745508291,
      "grad_norm": 0.12886938452720642,
      "learning_rate": 0.0004485586029984899,
      "loss": 1.5335720777511597,
      "step": 11776
    },
    {
      "epoch": 0.5557249087801949,
      "grad_norm": 0.12272178381681442,
      "learning_rate": 0.00042967849590597266,
      "loss": 1.536974310874939,
      "step": 12032
    },
    {
      "epoch": 0.5675488430095608,
      "grad_norm": 0.1320953667163849,
      "learning_rate": 0.0004109001650911621,
      "loss": 1.5350415706634521,
      "step": 12288
    },
    {
      "epoch": 0.5675488430095608,
      "eval_cos_loss": 0.4204618357224007,
      "eval_loss": 1.4744470647480934,
      "eval_mse_loss": 1.2642161459683283,
      "flow/cos_sim": 0.5795381778179238,
      "flow/improvement_ratio": 0.9622249373290093,
      "flow/mag_ratio_mean": 0.5585955846799563,
      "flow/mag_ratio_std": 0.21563580887366648,
      "step": 12288
    },
    {
      "epoch": 0.5675488430095608,
      "eval_cos_loss": 0.4204618357224007,
      "eval_loss": 1.4744470647480934,
      "eval_mse_loss": 1.2642161459683283,
      "eval_runtime": 39.1238,
      "eval_samples_per_second": 715.498,
      "eval_steps_per_second": 11.195,
      "flow/cos_sim": 0.5795381778179238,
      "flow/improvement_ratio": 0.9622249373290093,
      "flow/mag_ratio_mean": 0.5585955846799563,
      "flow/mag_ratio_std": 0.21563580887366648,
      "step": 12288
    },
    {
      "epoch": 0.5793727772389267,
      "grad_norm": 0.16382183134555817,
      "learning_rate": 0.0003922507884228551,
      "loss": 1.5361268520355225,
      "step": 12544
    },
    {
      "epoch": 0.5911967114682924,
      "grad_norm": 0.13758233189582825,
      "learning_rate": 0.00037375735713457723,
      "loss": 1.530663013458252,
      "step": 12800
    },
    {
      "epoch": 0.6030206456976583,
      "grad_norm": 0.13651148974895477,
      "learning_rate": 0.00035544663676018276,
      "loss": 1.528375506401062,
      "step": 13056
    },
    {
      "epoch": 0.6148445799270241,
      "grad_norm": 0.19377295672893524,
      "learning_rate": 0.00033734512839611255,
      "loss": 1.531029462814331,
      "step": 13312
    },
    {
      "epoch": 0.6148445799270241,
      "eval_cos_loss": 0.4196748673643696,
      "eval_loss": 1.4739454851302927,
      "eval_mse_loss": 1.2641080489986019,
      "flow/cos_sim": 0.5803251580832756,
      "flow/improvement_ratio": 0.9619928854241219,
      "flow/mag_ratio_mean": 0.5591708176756558,
      "flow/mag_ratio_std": 0.2099404629302896,
      "step": 13312
    },
    {
      "epoch": 0.6148445799270241,
      "eval_cos_loss": 0.4196748673643696,
      "eval_loss": 1.4739454851302927,
      "eval_mse_loss": 1.2641080489986019,
      "eval_runtime": 37.6325,
      "eval_samples_per_second": 743.851,
      "eval_steps_per_second": 11.639,
      "flow/cos_sim": 0.5803251580832756,
      "flow/improvement_ratio": 0.9619928854241219,
      "flow/mag_ratio_mean": 0.5591708176756558,
      "flow/mag_ratio_std": 0.2099404629302896,
      "step": 13312
    },
    {
      "epoch": 0.62666851415639,
      "grad_norm": 0.11970210075378418,
      "learning_rate": 0.0003194790303463687,
      "loss": 1.526605248451233,
      "step": 13568
    },
    {
      "epoch": 0.6384924483857558,
      "grad_norm": 0.16870331764221191,
      "learning_rate": 0.00030187420020572406,
      "loss": 1.5294699668884277,
      "step": 13824
    },
    {
      "epoch": 0.6503163826151217,
      "grad_norm": 0.20158159732818604,
      "learning_rate": 0.00028455611743603626,
      "loss": 1.5296140909194946,
      "step": 14080
    },
    {
      "epoch": 0.6621403168444876,
      "grad_norm": 0.17316782474517822,
      "learning_rate": 0.0002675498464898373,
      "loss": 1.5282411575317383,
      "step": 14336
    },
    {
      "epoch": 0.6621403168444876,
      "eval_cos_loss": 0.4197337330748501,
      "eval_loss": 1.4732592075927073,
      "eval_mse_loss": 1.263392340374864,
      "flow/cos_sim": 0.5802662803974326,
      "flow/improvement_ratio": 0.9630070868964609,
      "flow/mag_ratio_mean": 0.5561743245277231,
      "flow/mag_ratio_std": 0.21411516738536696,
      "step": 14336
    },
    {
      "epoch": 0.6621403168444876,
      "eval_cos_loss": 0.4197337330748501,
      "eval_loss": 1.4732592075927073,
      "eval_mse_loss": 1.263392340374864,
      "eval_runtime": 38.6362,
      "eval_samples_per_second": 724.527,
      "eval_steps_per_second": 11.337,
      "flow/cos_sim": 0.5802662803974326,
      "flow/improvement_ratio": 0.9630070868964609,
      "flow/mag_ratio_mean": 0.5561743245277231,
      "flow/mag_ratio_std": 0.21411516738536696,
      "step": 14336
    },
    {
      "epoch": 0.6739642510738534,
      "grad_norm": 0.1980622410774231,
      "learning_rate": 0.0002508800005345623,
      "loss": 1.527830719947815,
      "step": 14592
    },
    {
      "epoch": 0.6857881853032193,
      "grad_norm": 0.13991132378578186,
      "learning_rate": 0.00023457070582992562,
      "loss": 1.5318052768707275,
      "step": 14848
    },
    {
      "epoch": 0.6976121195325851,
      "grad_norm": 0.149434432387352,
      "learning_rate": 0.00021864556680999692,
      "loss": 1.5274654626846313,
      "step": 15104
    },
    {
      "epoch": 0.709436053761951,
      "grad_norm": 0.13668516278266907,
      "learning_rate": 0.0002031276319205152,
      "loss": 1.5278632640838623,
      "step": 15360
    },
    {
      "epoch": 0.709436053761951,
      "eval_cos_loss": 0.41798419257005054,
      "eval_loss": 1.4683536525186338,
      "eval_mse_loss": 1.259361555315044,
      "flow/cos_sim": 0.5820158255971186,
      "flow/improvement_ratio": 0.9633506106459386,
      "flow/mag_ratio_mean": 0.5602124498042886,
      "flow/mag_ratio_std": 0.21119761075739446,
      "step": 15360
    },
    {
      "epoch": 0.709436053761951,
      "eval_cos_loss": 0.41798419257005054,
      "eval_loss": 1.4683536525186338,
      "eval_mse_loss": 1.259361555315044,
      "eval_runtime": 37.7713,
      "eval_samples_per_second": 741.119,
      "eval_steps_per_second": 11.596,
      "flow/cos_sim": 0.5820158255971186,
      "flow/improvement_ratio": 0.9633506106459386,
      "flow/mag_ratio_mean": 0.5602124498042886,
      "flow/mag_ratio_std": 0.21119761075739446,
      "step": 15360
    },
    {
      "epoch": 0.7212599879913169,
      "grad_norm": 0.10699856281280518,
      "learning_rate": 0.00018803936026088542,
      "loss": 1.5269209146499634,
      "step": 15616
    },
    {
      "epoch": 0.7330839222206826,
      "grad_norm": 0.14415434002876282,
      "learning_rate": 0.00017340258907913464,
      "loss": 1.5266590118408203,
      "step": 15872
    },
    {
      "epoch": 0.7449078564500485,
      "grad_norm": 0.10905779153108597,
      "learning_rate": 0.0001592385021668743,
      "loss": 1.524775505065918,
      "step": 16128
    },
    {
      "epoch": 0.7567317906794143,
      "grad_norm": 0.19332484900951385,
      "learning_rate": 0.0001455675992000087,
      "loss": 1.528196096420288,
      "step": 16384
    },
    {
      "epoch": 0.7567317906794143,
      "eval_cos_loss": 0.4177281832041806,
      "eval_loss": 1.4672707036205623,
      "eval_mse_loss": 1.258406613243225,
      "flow/cos_sim": 0.5822718332619428,
      "flow/improvement_ratio": 0.9624373912811279,
      "flow/mag_ratio_mean": 0.5578821809052332,
      "flow/mag_ratio_std": 0.21358931578321544,
      "step": 16384
    },
    {
      "epoch": 0.7567317906794143,
      "eval_cos_loss": 0.4177281832041806,
      "eval_loss": 1.4672707036205623,
      "eval_mse_loss": 1.258406613243225,
      "eval_runtime": 38.8935,
      "eval_samples_per_second": 719.735,
      "eval_steps_per_second": 11.262,
      "flow/cos_sim": 0.5822718332619428,
      "flow/improvement_ratio": 0.9624373912811279,
      "flow/mag_ratio_mean": 0.5578821809052332,
      "flow/mag_ratio_std": 0.21358931578321544,
      "step": 16384
    },
    {
      "epoch": 0.7685557249087802,
      "grad_norm": 0.17042651772499084,
      "learning_rate": 0.000132409666069565,
      "loss": 1.524718999862671,
      "step": 16640
    },
    {
      "epoch": 0.780379659138146,
      "grad_norm": 0.18739284574985504,
      "learning_rate": 0.0001197837462455823,
      "loss": 1.5236212015151978,
      "step": 16896
    },
    {
      "epoch": 0.7922035933675119,
      "grad_norm": 0.18229515850543976,
      "learning_rate": 0.00010770811321550749,
      "loss": 1.5214073657989502,
      "step": 17152
    },
    {
      "epoch": 0.8040275275968778,
      "grad_norm": 0.1200081929564476,
      "learning_rate": 9.620024403698591e-05,
      "loss": 1.5241215229034424,
      "step": 17408
    },
    {
      "epoch": 0.8040275275968778,
      "eval_cos_loss": 0.41774995141922067,
      "eval_loss": 1.4673981081405187,
      "eval_mse_loss": 1.2585231351525816,
      "flow/cos_sim": 0.5822500714428349,
      "flow/improvement_ratio": 0.9632431905563563,
      "flow/mag_ratio_mean": 0.5575322135942712,
      "flow/mag_ratio_std": 0.2133726158506794,
      "step": 17408
    },
    {
      "epoch": 0.8040275275968778,
      "eval_cos_loss": 0.41774995141922067,
      "eval_loss": 1.4673981081405187,
      "eval_mse_loss": 1.2585231351525816,
      "eval_runtime": 38.8056,
      "eval_samples_per_second": 721.365,
      "eval_steps_per_second": 11.287,
      "flow/cos_sim": 0.5822500714428349,
      "flow/improvement_ratio": 0.9632431905563563,
      "flow/mag_ratio_mean": 0.5575322135942712,
      "flow/mag_ratio_std": 0.2133726158506794,
      "step": 17408
    },
    {
      "epoch": 0.8158514618262436,
      "grad_norm": 0.20400674641132355,
      "learning_rate": 8.527679404332429e-05,
      "loss": 1.527288794517517,
      "step": 17664
    },
    {
      "epoch": 0.8276753960556095,
      "grad_norm": 0.13825097680091858,
      "learning_rate": 7.495357273823544e-05,
      "loss": 1.5233813524246216,
      "step": 17920
    },
    {
      "epoch": 0.8394993302849753,
      "grad_norm": 0.1886366605758667,
      "learning_rate": 6.524552091475183e-05,
      "loss": 1.5250685214996338,
      "step": 18176
    },
    {
      "epoch": 0.8513232645143411,
      "grad_norm": 0.12399227172136307,
      "learning_rate": 5.6166689031422024e-05,
      "loss": 1.5211102962493896,
      "step": 18432
    },
    {
      "epoch": 0.8513232645143411,
      "eval_cos_loss": 0.4172959377368291,
      "eval_loss": 1.467090511975223,
      "eval_mse_loss": 1.2584425412356581,
      "flow/cos_sim": 0.5827040846489336,
      "flow/improvement_ratio": 0.9641230167591408,
      "flow/mag_ratio_mean": 0.559296594771076,
      "flow/mag_ratio_std": 0.2129400044203349,
      "step": 18432
    },
    {
      "epoch": 0.8513232645143411,
      "eval_cos_loss": 0.4172959377368291,
      "eval_loss": 1.467090511975223,
      "eval_mse_loss": 1.2584425412356581,
      "eval_runtime": 38.949,
      "eval_samples_per_second": 718.708,
      "eval_steps_per_second": 11.245,
      "flow/cos_sim": 0.5827040846489336,
      "flow/improvement_ratio": 0.9641230167591408,
      "flow/mag_ratio_mean": 0.559296594771076,
      "flow/mag_ratio_std": 0.2129400044203349,
      "step": 18432
    },
    {
      "epoch": 0.8631471987437069,
      "grad_norm": 0.15448303520679474,
      "learning_rate": 4.773021687709067e-05,
      "loss": 1.524503231048584,
      "step": 18688
    },
    {
      "epoch": 0.8749711329730728,
      "grad_norm": 0.11015952378511429,
      "learning_rate": 3.994831455368719e-05,
      "loss": 1.5246198177337646,
      "step": 18944
    },
    {
      "epoch": 0.8867950672024387,
      "grad_norm": 0.14956603944301605,
      "learning_rate": 3.283224480455282e-05,
      "loss": 1.524648666381836,
      "step": 19200
    },
    {
      "epoch": 0.8986190014318045,
      "grad_norm": 0.1354081630706787,
      "learning_rate": 2.639230671387627e-05,
      "loss": 1.5228582620620728,
      "step": 19456
    },
    {
      "epoch": 0.8986190014318045,
      "eval_cos_loss": 0.4183324677486942,
      "eval_loss": 1.4691238653714254,
      "eval_mse_loss": 1.2599576338785423,
      "flow/cos_sim": 0.5816675356533973,
      "flow/improvement_ratio": 0.9630278509501453,
      "flow/mag_ratio_mean": 0.5622789722327228,
      "flow/mag_ratio_std": 0.21166673544198955,
      "step": 19456
    },
    {
      "epoch": 0.8986190014318045,
      "eval_cos_loss": 0.4183324677486942,
      "eval_loss": 1.4691238653714254,
      "eval_mse_loss": 1.2599576338785423,
      "eval_runtime": 38.8176,
      "eval_samples_per_second": 721.141,
      "eval_steps_per_second": 11.284,
      "flow/cos_sim": 0.5816675356533973,
      "flow/improvement_ratio": 0.9630278509501453,
      "flow/mag_ratio_mean": 0.5622789722327228,
      "flow/mag_ratio_std": 0.21166673544198955,
      "step": 19456
    },
    {
      "epoch": 0.9104429356611704,
      "grad_norm": 0.11552918702363968,
      "learning_rate": 2.063782080083576e-05,
      "loss": 1.5217269659042358,
      "step": 19712
    },
    {
      "epoch": 0.9222668698905362,
      "grad_norm": 0.14077740907669067,
      "learning_rate": 1.557711553001523e-05,
      "loss": 1.5255944728851318,
      "step": 19968
    },
    {
      "epoch": 0.9340908041199021,
      "grad_norm": 0.11086593568325043,
      "learning_rate": 1.1217515257622269e-05,
      "loss": 1.5237233638763428,
      "step": 20224
    },
    {
      "epoch": 0.945914738349268,
      "grad_norm": 0.1408848911523819,
      "learning_rate": 7.565329630950746e-06,
      "loss": 1.5244028568267822,
      "step": 20480
    },
    {
      "epoch": 0.945914738349268,
      "eval_cos_loss": 0.41753342645625546,
      "eval_loss": 1.4667748912284364,
      "eval_mse_loss": 1.2580081817766302,
      "flow/cos_sim": 0.5824665990594315,
      "flow/improvement_ratio": 0.9622961689620257,
      "flow/mag_ratio_mean": 0.5584562299730571,
      "flow/mag_ratio_std": 0.2129541815483951,
      "step": 20480
    },
    {
      "epoch": 0.945914738349268,
      "eval_cos_loss": 0.41753342645625546,
      "eval_loss": 1.4667748912284364,
      "eval_mse_loss": 1.2580081817766302,
      "eval_runtime": 38.5143,
      "eval_samples_per_second": 726.82,
      "eval_steps_per_second": 11.372,
      "flow/cos_sim": 0.5824665990594315,
      "flow/improvement_ratio": 0.9622961689620257,
      "flow/mag_ratio_mean": 0.5584562299730571,
      "flow/mag_ratio_std": 0.2129541815483951,
      "step": 20480
    },
    {
      "epoch": 0.9577386725786338,
      "grad_norm": 0.11133582890033722,
      "learning_rate": 4.62584445643166e-06,
      "loss": 1.5237617492675781,
      "step": 20736
    },
    {
      "epoch": 0.9695626068079997,
      "grad_norm": 0.13919079303741455,
      "learning_rate": 2.40331404948807e-06,
      "loss": 1.5243273973464966,
      "step": 20992
    },
    {
      "epoch": 0.9813865410373654,
      "grad_norm": 0.10569830238819122,
      "learning_rate": 9.009550772663965e-07,
      "loss": 1.524424433708191,
      "step": 21248
    },
    {
      "epoch": 0.9932104752667313,
      "grad_norm": 0.10687630623579025,
      "learning_rate": 1.2094190315575791e-07,
      "loss": 1.5235631465911865,
      "step": 21504
    },
    {
      "epoch": 0.9932104752667313,
      "eval_cos_loss": 0.4184109731896283,
      "eval_loss": 1.4685546744904017,
      "eval_mse_loss": 1.25934919020901,
      "flow/cos_sim": 0.5815890253134514,
      "flow/improvement_ratio": 0.9625635148455564,
      "flow/mag_ratio_mean": 0.5583232171731453,
      "flow/mag_ratio_std": 0.2131242363216126,
      "step": 21504
    },
    {
      "epoch": 0.9932104752667313,
      "eval_cos_loss": 0.4184109731896283,
      "eval_loss": 1.4685546744904017,
      "eval_mse_loss": 1.25934919020901,
      "eval_runtime": 37.8094,
      "eval_samples_per_second": 740.371,
      "eval_steps_per_second": 11.584,
      "flow/cos_sim": 0.5815890253134514,
      "flow/improvement_ratio": 0.9625635148455564,
      "flow/mag_ratio_mean": 0.5583232171731453,
      "flow/mag_ratio_std": 0.2131242363216126,
      "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
}