{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.7567317906794143,
  "eval_steps": 1024,
  "global_step": 16384,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.011823934229365849,
      "grad_norm": 0.528586745262146,
      "learning_rate": 4.9804687500000004e-05,
      "loss": 0.942461371421814,
      "step": 256
    },
    {
      "epoch": 0.023647868458731697,
      "grad_norm": 3.552647352218628,
      "learning_rate": 9.98046875e-05,
      "loss": 0.5787786245346069,
      "step": 512
    },
    {
      "epoch": 0.03547180268809755,
      "grad_norm": 3.1322181224823,
      "learning_rate": 9.99640996023194e-05,
      "loss": 0.4046097695827484,
      "step": 768
    },
    {
      "epoch": 0.047295736917463395,
      "grad_norm": 3.1972620487213135,
      "learning_rate": 9.985588674043959e-05,
      "loss": 0.3494558036327362,
      "step": 1024
    },
    {
      "epoch": 0.047295736917463395,
      "eval_cos_loss": 0.12621666669641454,
      "eval_loss": 0.2689284305662325,
      "eval_mse_loss": 0.2689284305662325,
      "flow/cos_sim": 0.873783351487765,
      "flow/improvement_ratio": 0.9758892705723575,
      "flow/mag_ratio_mean": 0.7793661633582964,
      "flow/mag_ratio_std": 0.19252585356025936,
      "step": 1024
    },
    {
      "epoch": 0.047295736917463395,
      "eval_cos_loss": 0.12621666669641454,
      "eval_loss": 0.2689284305662325,
      "eval_mse_loss": 0.2689284305662325,
      "eval_runtime": 38.7216,
      "eval_samples_per_second": 722.93,
      "eval_steps_per_second": 11.312,
      "flow/cos_sim": 0.873783351487765,
      "flow/improvement_ratio": 0.9758892705723575,
      "flow/mag_ratio_mean": 0.7793661633582964,
      "flow/mag_ratio_std": 0.19252585356025936,
      "step": 1024
    },
    {
      "epoch": 0.05911967114682925,
      "grad_norm": 2.4856488704681396,
      "learning_rate": 9.967551747861388e-05,
      "loss": 0.32386085391044617,
      "step": 1280
    },
    {
      "epoch": 0.0709436053761951,
      "grad_norm": 3.6864614486694336,
      "learning_rate": 9.94232528651847e-05,
      "loss": 0.305107444524765,
      "step": 1536
    },
    {
      "epoch": 0.08276753960556095,
      "grad_norm": 2.9081523418426514,
      "learning_rate": 9.909945800260091e-05,
      "loss": 0.29414206743240356,
      "step": 1792
    },
    {
      "epoch": 0.09459147383492679,
      "grad_norm": 4.508695602416992,
      "learning_rate": 9.870460151900524e-05,
      "loss": 0.2828270196914673,
      "step": 2048
    },
    {
      "epoch": 0.09459147383492679,
      "eval_cos_loss": 0.09525654536404021,
      "eval_loss": 0.21123698139435623,
      "eval_mse_loss": 0.21123698139435623,
      "flow/cos_sim": 0.9047434700134138,
      "flow/improvement_ratio": 0.982480168070423,
      "flow/mag_ratio_mean": 0.8272033925742319,
      "flow/mag_ratio_std": 0.16838706260946787,
      "step": 2048
    },
    {
      "epoch": 0.09459147383492679,
      "eval_cos_loss": 0.09525654536404021,
      "eval_loss": 0.21123698139435623,
      "eval_mse_loss": 0.21123698139435623,
      "eval_runtime": 38.5519,
      "eval_samples_per_second": 726.113,
      "eval_steps_per_second": 11.361,
      "flow/cos_sim": 0.9047434700134138,
      "flow/improvement_ratio": 0.982480168070423,
      "flow/mag_ratio_mean": 0.8272033925742319,
      "flow/mag_ratio_std": 0.16838706260946787,
      "step": 2048
    },
    {
      "epoch": 0.10641540806429264,
      "grad_norm": 4.400907039642334,
      "learning_rate": 9.823925488998887e-05,
      "loss": 0.2744397222995758,
      "step": 2304
    },
    {
      "epoch": 0.1182393422936585,
      "grad_norm": 4.855555534362793,
      "learning_rate": 9.770409161149526e-05,
      "loss": 0.2671351730823517,
      "step": 2560
    },
    {
      "epoch": 0.13006327652302435,
      "grad_norm": 5.30946159362793,
      "learning_rate": 9.709988622506974e-05,
      "loss": 0.26279500126838684,
      "step": 2816
    },
    {
      "epoch": 0.1418872107523902,
      "grad_norm": 3.811448097229004,
      "learning_rate": 9.642751319686591e-05,
      "loss": 0.25749024748802185,
      "step": 3072
    },
    {
      "epoch": 0.1418872107523902,
      "eval_cos_loss": 0.08731762273024478,
      "eval_loss": 0.18876669879101182,
      "eval_mse_loss": 0.18876669879101182,
      "flow/cos_sim": 0.9126823925536517,
      "flow/improvement_ratio": 0.9797838105730814,
      "flow/mag_ratio_mean": 0.858971359795087,
      "flow/mag_ratio_std": 0.14750978925592823,
      "step": 3072
    },
    {
      "epoch": 0.1418872107523902,
      "eval_cos_loss": 0.08731762273024478,
      "eval_loss": 0.18876669879101182,
      "eval_mse_loss": 0.18876669879101182,
      "eval_runtime": 39.5006,
      "eval_samples_per_second": 708.672,
      "eval_steps_per_second": 11.088,
      "flow/cos_sim": 0.9126823925536517,
      "flow/improvement_ratio": 0.9797838105730814,
      "flow/mag_ratio_mean": 0.858971359795087,
      "flow/mag_ratio_std": 0.14750978925592823,
      "step": 3072
    },
    {
      "epoch": 0.15371114498175603,
      "grad_norm": 4.319483757019043,
      "learning_rate": 9.568794565203123e-05,
      "loss": 0.2530311346054077,
      "step": 3328
    },
    {
      "epoch": 0.1655350792111219,
      "grad_norm": 6.430596828460693,
      "learning_rate": 9.488225396630348e-05,
      "loss": 0.24780720472335815,
      "step": 3584
    },
    {
      "epoch": 0.17735901344048774,
      "grad_norm": 5.565337181091309,
      "learning_rate": 9.401160421685646e-05,
      "loss": 0.24508802592754364,
      "step": 3840
    },
    {
      "epoch": 0.18918294766985358,
      "grad_norm": 4.396243095397949,
      "learning_rate": 9.307725649463714e-05,
      "loss": 0.24013251066207886,
      "step": 4096
    },
    {
      "epoch": 0.18918294766985358,
      "eval_cos_loss": 0.07958807458582262,
      "eval_loss": 0.17615001155361193,
      "eval_mse_loss": 0.17615001155361193,
      "flow/cos_sim": 0.9204119443893433,
      "flow/improvement_ratio": 0.9829439191513409,
      "flow/mag_ratio_mean": 0.8636429496279591,
      "flow/mag_ratio_std": 0.1477059963325115,
      "step": 4096
    },
    {
      "epoch": 0.18918294766985358,
      "eval_cos_loss": 0.07958807458582262,
      "eval_loss": 0.17615001155361193,
      "eval_mse_loss": 0.17615001155361193,
      "eval_runtime": 39.223,
      "eval_samples_per_second": 713.688,
      "eval_steps_per_second": 11.167,
      "flow/cos_sim": 0.9204119443893433,
      "flow/improvement_ratio": 0.9829439191513409,
      "flow/mag_ratio_mean": 0.8636429496279591,
      "flow/mag_ratio_std": 0.1477059963325115,
      "step": 4096
    },
    {
      "epoch": 0.20100688189921945,
      "grad_norm": 5.331676483154297,
      "learning_rate": 9.20805630806366e-05,
      "loss": 0.23666875064373016,
      "step": 4352
    },
    {
      "epoch": 0.2128308161285853,
      "grad_norm": 5.4501800537109375,
      "learning_rate": 9.102296648873445e-05,
      "loss": 0.2347797006368637,
      "step": 4608
    },
    {
      "epoch": 0.22465475035795113,
      "grad_norm": 7.467021465301514,
      "learning_rate": 8.990599737794927e-05,
      "loss": 0.23136268556118011,
      "step": 4864
    },
    {
      "epoch": 0.236478684587317,
      "grad_norm": 5.311818599700928,
      "learning_rate": 8.873127233711644e-05,
      "loss": 0.2264125496149063,
      "step": 5120
    },
    {
      "epoch": 0.236478684587317,
      "eval_cos_loss": 0.07416401562795519,
      "eval_loss": 0.16512389409623734,
      "eval_mse_loss": 0.16512389409623734,
      "flow/cos_sim": 0.9258360114271782,
      "flow/improvement_ratio": 0.9846288648914528,
      "flow/mag_ratio_mean": 0.8722622129198623,
      "flow/mag_ratio_std": 0.1423258438672377,
      "step": 5120
    },
    {
      "epoch": 0.236478684587317,
      "eval_cos_loss": 0.07416401562795519,
      "eval_loss": 0.16512389409623734,
      "eval_mse_loss": 0.16512389409623734,
      "eval_runtime": 40.461,
      "eval_samples_per_second": 691.851,
      "eval_steps_per_second": 10.825,
      "flow/cos_sim": 0.9258360114271782,
      "flow/improvement_ratio": 0.9846288648914528,
      "flow/mag_ratio_mean": 0.8722622129198623,
      "flow/mag_ratio_std": 0.1423258438672377,
      "step": 5120
    },
    {
      "epoch": 0.24830261881668284,
      "grad_norm": 5.7909674644470215,
      "learning_rate": 8.750049154520012e-05,
      "loss": 0.22343386709690094,
      "step": 5376
    },
    {
      "epoch": 0.2601265530460487,
      "grad_norm": 4.872880458831787,
      "learning_rate": 8.621543631062488e-05,
      "loss": 0.2214333564043045,
      "step": 5632
    },
    {
      "epoch": 0.27195048727541454,
      "grad_norm": 8.654791831970215,
      "learning_rate": 8.487796649318904e-05,
      "loss": 0.216692253947258,
      "step": 5888
    },
    {
      "epoch": 0.2837744215047804,
      "grad_norm": 4.427462100982666,
      "learning_rate": 8.349001781229053e-05,
      "loss": 0.21447119116783142,
      "step": 6144
    },
    {
      "epoch": 0.2837744215047804,
      "eval_cos_loss": 0.0701922818797228,
      "eval_loss": 0.1572090190363257,
      "eval_mse_loss": 0.1572090190363257,
      "flow/cos_sim": 0.9298077467914041,
      "flow/improvement_ratio": 0.9808430573711656,
      "flow/mag_ratio_mean": 0.8736172293963498,
      "flow/mag_ratio_std": 0.13489851995145893,
      "step": 6144
    },
    {
      "epoch": 0.2837744215047804,
      "eval_cos_loss": 0.0701922818797228,
      "eval_loss": 0.1572090190363257,
      "eval_mse_loss": 0.1572090190363257,
      "eval_runtime": 39.4187,
      "eval_samples_per_second": 710.145,
      "eval_steps_per_second": 11.111,
      "flow/cos_sim": 0.9298077467914041,
      "flow/improvement_ratio": 0.9808430573711656,
      "flow/mag_ratio_mean": 0.8736172293963498,
      "flow/mag_ratio_std": 0.13489851995145893,
      "step": 6144
    },
    {
      "epoch": 0.2955983557341462,
      "grad_norm": 3.5116236209869385,
      "learning_rate": 8.205359904536107e-05,
      "loss": 0.21204149723052979,
      "step": 6400
    },
    {
      "epoch": 0.30742228996351206,
      "grad_norm": 4.563979625701904,
      "learning_rate": 8.057078912056364e-05,
      "loss": 0.21015891432762146,
      "step": 6656
    },
    {
      "epoch": 0.3192462241928779,
      "grad_norm": 4.94021463394165,
      "learning_rate": 7.904373410796087e-05,
      "loss": 0.20794054865837097,
      "step": 6912
    },
    {
      "epoch": 0.3310701584222438,
      "grad_norm": 4.7865891456604,
      "learning_rate": 7.747464411350877e-05,
      "loss": 0.2068762332201004,
      "step": 7168
    },
    {
      "epoch": 0.3310701584222438,
      "eval_cos_loss": 0.06751056726528629,
      "eval_loss": 0.15150919330378645,
      "eval_mse_loss": 0.15150919330378645,
      "flow/cos_sim": 0.9324894422265493,
      "flow/improvement_ratio": 0.9829098535999315,
      "flow/mag_ratio_mean": 0.8796486472158127,
      "flow/mag_ratio_std": 0.13032293935480727,
      "step": 7168
    },
    {
      "epoch": 0.3310701584222438,
      "eval_cos_loss": 0.06751056726528629,
      "eval_loss": 0.15150919330378645,
      "eval_mse_loss": 0.15150919330378645,
      "eval_runtime": 38.541,
      "eval_samples_per_second": 726.317,
      "eval_steps_per_second": 11.365,
      "flow/cos_sim": 0.9324894422265493,
      "flow/improvement_ratio": 0.9829098535999315,
      "flow/mag_ratio_mean": 0.8796486472158127,
      "flow/mag_ratio_std": 0.13032293935480727,
      "step": 7168
    },
    {
      "epoch": 0.34289409265160964,
      "grad_norm": 6.607056617736816,
      "learning_rate": 7.58657900803716e-05,
      "loss": 0.2057526409626007,
      "step": 7424
    },
    {
      "epoch": 0.3547180268809755,
      "grad_norm": 3.4724538326263428,
      "learning_rate": 7.42195005021869e-05,
      "loss": 0.2034897804260254,
      "step": 7680
    },
    {
      "epoch": 0.3665419611103413,
      "grad_norm": 3.786540985107422,
      "learning_rate": 7.253815805303786e-05,
      "loss": 0.20293587446212769,
      "step": 7936
    },
    {
      "epoch": 0.37836589533970716,
      "grad_norm": 5.45916223526001,
      "learning_rate": 7.082419613901028e-05,
      "loss": 0.20195142924785614,
      "step": 8192
    },
    {
      "epoch": 0.37836589533970716,
      "eval_cos_loss": 0.06593506325665674,
      "eval_loss": 0.14644465646515153,
      "eval_mse_loss": 0.14644465646515153,
      "flow/cos_sim": 0.9340649522058496,
      "flow/improvement_ratio": 0.9748687266486965,
      "flow/mag_ratio_mean": 0.8922792132858817,
      "flow/mag_ratio_std": 0.12403852046896878,
      "step": 8192
    },
    {
      "epoch": 0.37836589533970716,
      "eval_cos_loss": 0.06593506325665674,
      "eval_loss": 0.14644465646515153,
      "eval_mse_loss": 0.14644465646515153,
      "eval_runtime": 37.6677,
      "eval_samples_per_second": 743.157,
      "eval_steps_per_second": 11.628,
      "flow/cos_sim": 0.9340649522058496,
      "flow/improvement_ratio": 0.9748687266486965,
      "flow/mag_ratio_mean": 0.8922792132858817,
      "flow/mag_ratio_std": 0.12403852046896878,
      "step": 8192
    },
    {
      "epoch": 0.390189829569073,
      "grad_norm": 4.096855163574219,
      "learning_rate": 6.908009537632513e-05,
      "loss": 0.19950586557388306,
      "step": 8448
    },
    {
      "epoch": 0.4020137637984389,
      "grad_norm": 5.261643409729004,
      "learning_rate": 6.730838000114404e-05,
      "loss": 0.20048308372497559,
      "step": 8704
    },
    {
      "epoch": 0.41383769802780473,
      "grad_norm": 5.638973236083984,
      "learning_rate": 6.551161421624341e-05,
      "loss": 0.19841572642326355,
      "step": 8960
    },
    {
      "epoch": 0.4256616322571706,
      "grad_norm": 4.699023246765137,
      "learning_rate": 6.369239847984518e-05,
      "loss": 0.19725315272808075,
      "step": 9216
    },
    {
      "epoch": 0.4256616322571706,
      "eval_cos_loss": 0.06321680692947346,
      "eval_loss": 0.14280268150173367,
      "eval_mse_loss": 0.14280268150173367,
      "flow/cos_sim": 0.9367832058368752,
      "flow/improvement_ratio": 0.9784972767035166,
      "flow/mag_ratio_mean": 0.8909378552545696,
      "flow/mag_ratio_std": 0.12450728342617483,
      "step": 9216
    },
    {
      "epoch": 0.4256616322571706,
      "eval_cos_loss": 0.06321680692947346,
      "eval_loss": 0.14280268150173367,
      "eval_mse_loss": 0.14280268150173367,
      "eval_runtime": 37.7105,
      "eval_samples_per_second": 742.313,
      "eval_steps_per_second": 11.615,
      "flow/cos_sim": 0.9367832058368752,
      "flow/improvement_ratio": 0.9784972767035166,
      "flow/mag_ratio_mean": 0.8909378552545696,
      "flow/mag_ratio_std": 0.12450728342617483,
      "step": 9216
    },
    {
      "epoch": 0.4374855664865364,
      "grad_norm": 5.382142543792725,
      "learning_rate": 6.185336574197478e-05,
      "loss": 0.19487647712230682,
      "step": 9472
    },
    {
      "epoch": 0.44930950071590225,
      "grad_norm": 5.455441951751709,
      "learning_rate": 5.999717763379407e-05,
      "loss": 0.1949186772108078,
      "step": 9728
    },
    {
      "epoch": 0.4611334349452681,
      "grad_norm": 3.58408522605896,
      "learning_rate": 5.812652061542364e-05,
      "loss": 0.19281470775604248,
      "step": 9984
    },
    {
      "epoch": 0.472957369174634,
      "grad_norm": 7.520720481872559,
      "learning_rate": 5.624410208783071e-05,
      "loss": 0.19212757050991058,
      "step": 10240
    },
    {
      "epoch": 0.472957369174634,
      "eval_cos_loss": 0.061764058457117646,
      "eval_loss": 0.1391820370912824,
      "eval_mse_loss": 0.1391820370912824,
      "flow/cos_sim": 0.9382359673987785,
      "flow/improvement_ratio": 0.9751606145405878,
      "flow/mag_ratio_mean": 0.899074621804773,
      "flow/mag_ratio_std": 0.12071822867750033,
      "step": 10240
    },
    {
      "epoch": 0.472957369174634,
      "eval_cos_loss": 0.061764058457117646,
      "eval_loss": 0.1391820370912824,
      "eval_mse_loss": 0.1391820370912824,
      "eval_runtime": 37.8004,
      "eval_samples_per_second": 740.548,
      "eval_steps_per_second": 11.587,
      "flow/cos_sim": 0.9382359673987785,
      "flow/improvement_ratio": 0.9751606145405878,
      "flow/mag_ratio_mean": 0.899074621804773,
      "flow/mag_ratio_std": 0.12071822867750033,
      "step": 10240
    },
    {
      "epoch": 0.48478130340399983,
      "grad_norm": 6.382692813873291,
      "learning_rate": 5.4352646474408806e-05,
      "loss": 0.1926891803741455,
      "step": 10496
    },
    {
      "epoch": 0.49660523763336567,
      "grad_norm": 4.348917007446289,
      "learning_rate": 5.24548912779213e-05,
      "loss": 0.19107094407081604,
      "step": 10752
    },
    {
      "epoch": 0.5084291718627315,
      "grad_norm": 4.5084733963012695,
      "learning_rate": 5.055358311851499e-05,
      "loss": 0.18951745331287384,
      "step": 11008
    },
    {
      "epoch": 0.5202531060920974,
      "grad_norm": 7.585500240325928,
      "learning_rate": 4.8651473758538116e-05,
      "loss": 0.18960337340831757,
      "step": 11264
    },
    {
      "epoch": 0.5202531060920974,
      "eval_cos_loss": 0.06116597219178938,
      "eval_loss": 0.1383266713147022,
      "eval_mse_loss": 0.1383266713147022,
      "flow/cos_sim": 0.9388340465282197,
      "flow/improvement_ratio": 0.9741401230091373,
      "flow/mag_ratio_mean": 0.894713885027524,
      "flow/mag_ratio_std": 0.12038007916108659,
      "step": 11264
    },
    {
      "epoch": 0.5202531060920974,
      "eval_cos_loss": 0.06116597219178938,
      "eval_loss": 0.1383266713147022,
      "eval_mse_loss": 0.1383266713147022,
      "eval_runtime": 37.8003,
      "eval_samples_per_second": 740.55,
      "eval_steps_per_second": 11.587,
      "flow/cos_sim": 0.9388340465282197,
      "flow/improvement_ratio": 0.9741401230091373,
      "flow/mag_ratio_mean": 0.894713885027524,
      "flow/mag_ratio_std": 0.12038007916108659,
      "step": 11264
    },
    {
      "epoch": 0.5320770403214632,
      "grad_norm": 6.697900295257568,
      "learning_rate": 4.675131611991607e-05,
      "loss": 0.1873694658279419,
      "step": 11520
    },
    {
      "epoch": 0.5439009745508291,
      "grad_norm": 4.439393520355225,
      "learning_rate": 4.485586029984899e-05,
      "loss": 0.18834149837493896,
      "step": 11776
    },
    {
      "epoch": 0.5557249087801949,
      "grad_norm": 4.207615852355957,
      "learning_rate": 4.2967849590597266e-05,
      "loss": 0.18644848465919495,
      "step": 12032
    },
    {
      "epoch": 0.5675488430095608,
      "grad_norm": 3.873199939727783,
      "learning_rate": 4.109001650911621e-05,
      "loss": 0.18585823476314545,
      "step": 12288
    },
    {
      "epoch": 0.5675488430095608,
      "eval_cos_loss": 0.059984874856322326,
      "eval_loss": 0.13662995146289808,
      "eval_mse_loss": 0.13662995146289808,
      "flow/cos_sim": 0.9400151448162724,
      "flow/improvement_ratio": 0.9759741743949994,
      "flow/mag_ratio_mean": 0.8886407714456184,
      "flow/mag_ratio_std": 0.11530450343677442,
      "step": 12288
    },
    {
      "epoch": 0.5675488430095608,
      "eval_cos_loss": 0.059984874856322326,
      "eval_loss": 0.13662995146289808,
      "eval_mse_loss": 0.13662995146289808,
      "eval_runtime": 37.6043,
      "eval_samples_per_second": 744.409,
      "eval_steps_per_second": 11.648,
      "flow/cos_sim": 0.9400151448162724,
      "flow/improvement_ratio": 0.9759741743949994,
      "flow/mag_ratio_mean": 0.8886407714456184,
      "flow/mag_ratio_std": 0.11530450343677442,
      "step": 12288
    },
    {
      "epoch": 0.5793727772389267,
      "grad_norm": 4.404060363769531,
      "learning_rate": 3.9225078842285515e-05,
      "loss": 0.18521401286125183,
      "step": 12544
    },
    {
      "epoch": 0.5911967114682924,
      "grad_norm": 4.842325210571289,
      "learning_rate": 3.7375735713457726e-05,
      "loss": 0.1849864274263382,
      "step": 12800
    },
    {
      "epoch": 0.6030206456976583,
      "grad_norm": 4.038843154907227,
      "learning_rate": 3.554466367601827e-05,
      "loss": 0.18520770967006683,
      "step": 13056
    },
    {
      "epoch": 0.6148445799270241,
      "grad_norm": 4.001939296722412,
      "learning_rate": 3.373451283961125e-05,
      "loss": 0.18334536254405975,
      "step": 13312
    },
    {
      "epoch": 0.6148445799270241,
      "eval_cos_loss": 0.05913267134939699,
      "eval_loss": 0.13375332260921122,
      "eval_mse_loss": 0.13375332260921122,
      "flow/cos_sim": 0.9408673460352911,
      "flow/improvement_ratio": 0.974089263236686,
      "flow/mag_ratio_mean": 0.899301437482442,
      "flow/mag_ratio_std": 0.11589696397792258,
      "step": 13312
    },
    {
      "epoch": 0.6148445799270241,
      "eval_cos_loss": 0.05913267134939699,
      "eval_loss": 0.13375332260921122,
      "eval_mse_loss": 0.13375332260921122,
      "eval_runtime": 37.5196,
      "eval_samples_per_second": 746.09,
      "eval_steps_per_second": 11.674,
      "flow/cos_sim": 0.9408673460352911,
      "flow/improvement_ratio": 0.974089263236686,
      "flow/mag_ratio_mean": 0.899301437482442,
      "flow/mag_ratio_std": 0.11589696397792258,
      "step": 13312
    },
    {
      "epoch": 0.62666851415639,
      "grad_norm": 5.12012243270874,
      "learning_rate": 3.194790303463687e-05,
      "loss": 0.18432201445102692,
      "step": 13568
    },
    {
      "epoch": 0.6384924483857558,
      "grad_norm": 8.441094398498535,
      "learning_rate": 3.0187420020572404e-05,
      "loss": 0.1838230937719345,
      "step": 13824
    },
    {
      "epoch": 0.6503163826151217,
      "grad_norm": 2.894239902496338,
      "learning_rate": 2.8455611743603627e-05,
      "loss": 0.1821696162223816,
      "step": 14080
    },
    {
      "epoch": 0.6621403168444876,
      "grad_norm": 4.168018341064453,
      "learning_rate": 2.675498464898373e-05,
      "loss": 0.182567298412323,
      "step": 14336
    },
    {
      "epoch": 0.6621403168444876,
      "eval_cos_loss": 0.05801974387507733,
      "eval_loss": 0.13219618526835963,
      "eval_mse_loss": 0.13219618526835963,
      "flow/cos_sim": 0.9419802757975173,
      "flow/improvement_ratio": 0.9763460176992634,
      "flow/mag_ratio_mean": 0.8963852926204193,
      "flow/mag_ratio_std": 0.11260453710272976,
      "step": 14336
    },
    {
      "epoch": 0.6621403168444876,
      "eval_cos_loss": 0.05801974387507733,
      "eval_loss": 0.13219618526835963,
      "eval_mse_loss": 0.13219618526835963,
      "eval_runtime": 37.7481,
      "eval_samples_per_second": 741.575,
      "eval_steps_per_second": 11.603,
      "flow/cos_sim": 0.9419802757975173,
      "flow/improvement_ratio": 0.9763460176992634,
      "flow/mag_ratio_mean": 0.8963852926204193,
      "flow/mag_ratio_std": 0.11260453710272976,
      "step": 14336
    },
    {
      "epoch": 0.6739642510738534,
      "grad_norm": 3.49924898147583,
      "learning_rate": 2.508800005345623e-05,
      "loss": 0.18235723674297333,
      "step": 14592
    },
    {
      "epoch": 0.6857881853032193,
      "grad_norm": 6.434615135192871,
      "learning_rate": 2.345707058299256e-05,
      "loss": 0.18260806798934937,
      "step": 14848
    },
    {
      "epoch": 0.6976121195325851,
      "grad_norm": 2.582023859024048,
      "learning_rate": 2.1864556680999694e-05,
      "loss": 0.18264462053775787,
      "step": 15104
    },
    {
      "epoch": 0.709436053761951,
      "grad_norm": 6.015897274017334,
      "learning_rate": 2.031276319205152e-05,
      "loss": 0.18182410299777985,
      "step": 15360
    },
    {
      "epoch": 0.709436053761951,
      "eval_cos_loss": 0.05760355804915025,
      "eval_loss": 0.13131338000705797,
      "eval_mse_loss": 0.13131338000705797,
      "flow/cos_sim": 0.9423964508047932,
      "flow/improvement_ratio": 0.9733326742638192,
      "flow/mag_ratio_mean": 0.901015834313005,
      "flow/mag_ratio_std": 0.11158135478899359,
      "step": 15360
    },
    {
      "epoch": 0.709436053761951,
      "eval_cos_loss": 0.05760355804915025,
      "eval_loss": 0.13131338000705797,
      "eval_mse_loss": 0.13131338000705797,
      "eval_runtime": 37.3381,
      "eval_samples_per_second": 749.716,
      "eval_steps_per_second": 11.731,
      "flow/cos_sim": 0.9423964508047932,
      "flow/improvement_ratio": 0.9733326742638192,
      "flow/mag_ratio_mean": 0.901015834313005,
      "flow/mag_ratio_std": 0.11158135478899359,
      "step": 15360
    },
    {
      "epoch": 0.7212599879913169,
      "grad_norm": 3.3396780490875244,
      "learning_rate": 1.8803936026088543e-05,
      "loss": 0.18102040886878967,
      "step": 15616
    },
    {
      "epoch": 0.7330839222206826,
      "grad_norm": 3.4187169075012207,
      "learning_rate": 1.7340258907913463e-05,
      "loss": 0.18094292283058167,
      "step": 15872
    },
    {
      "epoch": 0.7449078564500485,
      "grad_norm": 1.7081680297851562,
      "learning_rate": 1.592385021668743e-05,
      "loss": 0.18018104135990143,
      "step": 16128
    },
    {
      "epoch": 0.7567317906794143,
      "grad_norm": 6.856297969818115,
      "learning_rate": 1.4556759920000874e-05,
      "loss": 0.17991521954536438,
      "step": 16384
    },
    {
      "epoch": 0.7567317906794143,
      "eval_cos_loss": 0.057760640111279814,
      "eval_loss": 0.1307328896231303,
      "eval_mse_loss": 0.1307328896231303,
      "flow/cos_sim": 0.9422393834482045,
      "flow/improvement_ratio": 0.9711825578180078,
      "flow/mag_ratio_mean": 0.9053168751333402,
      "flow/mag_ratio_std": 0.11070215122000268,
      "step": 16384
    },
    {
      "epoch": 0.7567317906794143,
      "eval_cos_loss": 0.057760640111279814,
      "eval_loss": 0.1307328896231303,
      "eval_mse_loss": 0.1307328896231303,
      "eval_runtime": 37.0474,
      "eval_samples_per_second": 755.599,
      "eval_steps_per_second": 11.823,
      "flow/cos_sim": 0.9422393834482045,
      "flow/improvement_ratio": 0.9711825578180078,
      "flow/mag_ratio_mean": 0.9053168751333402,
      "flow/mag_ratio_std": 0.11070215122000268,
      "step": 16384
    }
  ],
  "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
}