{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.8986190014318045,
  "eval_steps": 1024,
  "global_step": 19456,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.047295736917463395,
      "grad_norm": 0.47292250394821167,
      "learning_rate": 0.00016650390625,
      "loss": 2.742399215698242,
      "step": 1024
    },
    {
      "epoch": 0.047295736917463395,
      "eval_cos_loss": 0.5961006330028517,
      "eval_loss": 2.449579068514854,
      "eval_mse_loss": 1.8534784379615088,
      "flow/cos_sim": 0.4038993830550207,
      "flow/improvement_ratio": 0.9530568593713247,
      "flow/mag_ratio_mean": 0.08299264077046146,
      "flow/mag_ratio_std": 0.04246563146307588,
      "step": 1024
    },
    {
      "epoch": 0.047295736917463395,
      "eval_cos_loss": 0.5961006330028517,
      "eval_loss": 2.449579068514854,
      "eval_mse_loss": 1.8534784379615088,
      "eval_runtime": 39.604,
      "eval_samples_per_second": 706.823,
      "eval_steps_per_second": 11.059,
      "flow/cos_sim": 0.4038993830550207,
      "flow/improvement_ratio": 0.9530568593713247,
      "flow/mag_ratio_mean": 0.08299264077046146,
      "flow/mag_ratio_std": 0.04246563146307588,
      "step": 1024
    },
    {
      "epoch": 0.09459147383492679,
      "grad_norm": 0.3767559826374054,
      "learning_rate": 0.0003331705729166667,
      "loss": 2.077958345413208,
      "step": 2048
    },
    {
      "epoch": 0.09459147383492679,
      "eval_cos_loss": 0.4873075728138832,
      "eval_loss": 1.9004751797680441,
      "eval_mse_loss": 1.413167606477868,
      "flow/cos_sim": 0.5126924207221427,
      "flow/improvement_ratio": 0.96191185848898,
      "flow/mag_ratio_mean": 0.5089414368753564,
      "flow/mag_ratio_std": 0.17938318715928353,
      "step": 2048
    },
    {
      "epoch": 0.09459147383492679,
      "eval_cos_loss": 0.4873075728138832,
      "eval_loss": 1.9004751797680441,
      "eval_mse_loss": 1.413167606477868,
      "eval_runtime": 40.5135,
      "eval_samples_per_second": 690.955,
      "eval_steps_per_second": 10.811,
      "flow/cos_sim": 0.5126924207221427,
      "flow/improvement_ratio": 0.96191185848898,
      "flow/mag_ratio_mean": 0.5089414368753564,
      "flow/mag_ratio_std": 0.17938318715928353,
      "step": 2048
    },
    {
      "epoch": 0.1418872107523902,
      "grad_norm": 0.32334470748901367,
      "learning_rate": 0.0004998372395833334,
      "loss": 1.8623729944229126,
      "step": 3072
    },
    {
      "epoch": 0.1418872107523902,
      "eval_cos_loss": 0.45344613587747423,
      "eval_loss": 1.7837548544417778,
      "eval_mse_loss": 1.3303087190405962,
      "flow/cos_sim": 0.5465538722195037,
      "flow/improvement_ratio": 0.9649395484086041,
      "flow/mag_ratio_mean": 0.545519895885633,
      "flow/mag_ratio_std": 0.2078168946722327,
      "step": 3072
    },
    {
      "epoch": 0.1418872107523902,
      "eval_cos_loss": 0.45344613587747423,
      "eval_loss": 1.7837548544417778,
      "eval_mse_loss": 1.3303087190405962,
      "eval_runtime": 39.4807,
      "eval_samples_per_second": 709.03,
      "eval_steps_per_second": 11.094,
      "flow/cos_sim": 0.5465538722195037,
      "flow/improvement_ratio": 0.9649395484086041,
      "flow/mag_ratio_mean": 0.545519895885633,
      "flow/mag_ratio_std": 0.2078168946722327,
      "step": 3072
    },
    {
      "epoch": 0.18918294766985358,
      "grad_norm": 0.3577847480773926,
      "learning_rate": 0.0004962689322628078,
      "loss": 1.775568962097168,
      "step": 4096
    },
    {
      "epoch": 0.18918294766985358,
      "eval_cos_loss": 0.4333892414967219,
      "eval_loss": 1.716203493342552,
      "eval_mse_loss": 1.282814250689119,
      "flow/cos_sim": 0.5666107895983953,
      "flow/improvement_ratio": 0.9660467817086608,
      "flow/mag_ratio_mean": 0.5606984513807515,
      "flow/mag_ratio_std": 0.21290324090822646,
      "step": 4096
    },
    {
      "epoch": 0.18918294766985358,
      "eval_cos_loss": 0.4333892414967219,
      "eval_loss": 1.716203493342552,
      "eval_mse_loss": 1.282814250689119,
      "eval_runtime": 39.672,
      "eval_samples_per_second": 705.612,
      "eval_steps_per_second": 11.041,
      "flow/cos_sim": 0.5666107895983953,
      "flow/improvement_ratio": 0.9660467817086608,
      "flow/mag_ratio_mean": 0.5606984513807515,
      "flow/mag_ratio_std": 0.21290324090822646,
      "step": 4096
    },
    {
      "epoch": 0.236478684587317,
      "grad_norm": 0.3629797697067261,
      "learning_rate": 0.00048517275746994595,
      "loss": 1.7335060834884644,
      "step": 5120
    },
    {
      "epoch": 0.236478684587317,
      "eval_cos_loss": 0.42527746058762345,
      "eval_loss": 1.6876028531218228,
      "eval_mse_loss": 1.2623253905609861,
      "flow/cos_sim": 0.5747225568991273,
      "flow/improvement_ratio": 0.9670584807385049,
      "flow/mag_ratio_mean": 0.569412647861324,
      "flow/mag_ratio_std": 0.22070125653727413,
      "step": 5120
    },
    {
      "epoch": 0.236478684587317,
      "eval_cos_loss": 0.42527746058762345,
      "eval_loss": 1.6876028531218228,
      "eval_mse_loss": 1.2623253905609861,
      "eval_runtime": 39.6097,
      "eval_samples_per_second": 706.721,
      "eval_steps_per_second": 11.058,
      "flow/cos_sim": 0.5747225568991273,
      "flow/improvement_ratio": 0.9670584807385049,
      "flow/mag_ratio_mean": 0.569412647861324,
      "flow/mag_ratio_std": 0.22070125653727413,
      "step": 5120
    },
    {
      "epoch": 0.2837744215047804,
      "grad_norm": 0.41703319549560547,
      "learning_rate": 0.0004670433228990193,
      "loss": 1.706570029258728,
      "step": 6144
    },
    {
      "epoch": 0.2837744215047804,
      "eval_cos_loss": 0.41543219704606216,
      "eval_loss": 1.6554871824778379,
      "eval_mse_loss": 1.2400549827101024,
      "flow/cos_sim": 0.5845678245912403,
      "flow/improvement_ratio": 0.9684235336573701,
      "flow/mag_ratio_mean": 0.580940374636759,
      "flow/mag_ratio_std": 0.21690591768314849,
      "step": 6144
    },
    {
      "epoch": 0.2837744215047804,
      "eval_cos_loss": 0.41543219704606216,
      "eval_loss": 1.6554871824778379,
      "eval_mse_loss": 1.2400549827101024,
      "eval_runtime": 39.4039,
      "eval_samples_per_second": 710.412,
      "eval_steps_per_second": 11.116,
      "flow/cos_sim": 0.5845678245912403,
      "flow/improvement_ratio": 0.9684235336573701,
      "flow/mag_ratio_mean": 0.580940374636759,
      "flow/mag_ratio_std": 0.21690591768314849,
      "step": 6144
    },
    {
      "epoch": 0.3310701584222438,
      "grad_norm": 0.4814172387123108,
      "learning_rate": 0.0004424228215503503,
      "loss": 1.6873632669448853,
      "step": 7168
    },
    {
      "epoch": 0.3310701584222438,
      "eval_cos_loss": 0.41339083641903585,
      "eval_loss": 1.6469332964997314,
      "eval_mse_loss": 1.2335424589239843,
      "flow/cos_sim": 0.5866091871914798,
      "flow/improvement_ratio": 0.9676676742290253,
      "flow/mag_ratio_mean": 0.5795940211102298,
      "flow/mag_ratio_std": 0.21971321344103442,
      "step": 7168
    },
    {
      "epoch": 0.3310701584222438,
      "eval_cos_loss": 0.41339083641903585,
      "eval_loss": 1.6469332964997314,
      "eval_mse_loss": 1.2335424589239843,
      "eval_runtime": 39.4159,
      "eval_samples_per_second": 710.196,
      "eval_steps_per_second": 11.112,
      "flow/cos_sim": 0.5866091871914798,
      "flow/improvement_ratio": 0.9676676742290253,
      "flow/mag_ratio_mean": 0.5795940211102298,
      "flow/mag_ratio_std": 0.21971321344103442,
      "step": 7168
    },
    {
      "epoch": 0.37836589533970716,
      "grad_norm": 0.3746698498725891,
      "learning_rate": 0.0004120475733264409,
      "loss": 1.6745531558990479,
      "step": 8192
    },
    {
      "epoch": 0.37836589533970716,
      "eval_cos_loss": 0.40882552949260903,
      "eval_loss": 1.6312648664870764,
      "eval_mse_loss": 1.2224393353614633,
      "flow/cos_sim": 0.5911744925529445,
      "flow/improvement_ratio": 0.9690862153763096,
      "flow/mag_ratio_mean": 0.5859045668007576,
      "flow/mag_ratio_std": 0.2201043140806564,
      "step": 8192
    },
    {
      "epoch": 0.37836589533970716,
      "eval_cos_loss": 0.40882552949260903,
      "eval_loss": 1.6312648664870764,
      "eval_mse_loss": 1.2224393353614633,
      "eval_runtime": 39.6627,
      "eval_samples_per_second": 705.777,
      "eval_steps_per_second": 11.043,
      "flow/cos_sim": 0.5911744925529445,
      "flow/improvement_ratio": 0.9690862153763096,
      "flow/mag_ratio_mean": 0.5859045668007576,
      "flow/mag_ratio_std": 0.2201043140806564,
      "step": 8192
    },
    {
      "epoch": 0.4256616322571706,
      "grad_norm": 0.47942957282066345,
      "learning_rate": 0.0003768624321203836,
      "loss": 1.6640042066574097,
      "step": 9216
    },
    {
      "epoch": 0.4256616322571706,
      "eval_cos_loss": 0.4063904314688896,
      "eval_loss": 1.6224081018199659,
      "eval_mse_loss": 1.2160176700108671,
      "flow/cos_sim": 0.593609596224136,
      "flow/improvement_ratio": 0.9695234370830397,
      "flow/mag_ratio_mean": 0.5825079079360178,
      "flow/mag_ratio_std": 0.21707500255407264,
      "step": 9216
    },
    {
      "epoch": 0.4256616322571706,
      "eval_cos_loss": 0.4063904314688896,
      "eval_loss": 1.6224081018199659,
      "eval_mse_loss": 1.2160176700108671,
      "eval_runtime": 39.7053,
      "eval_samples_per_second": 705.019,
      "eval_steps_per_second": 11.031,
      "flow/cos_sim": 0.593609596224136,
      "flow/improvement_ratio": 0.9695234370830397,
      "flow/mag_ratio_mean": 0.5825079079360178,
      "flow/mag_ratio_std": 0.21707500255407264,
      "step": 9216
    },
    {
      "epoch": 0.472957369174634,
      "grad_norm": 0.35636550188064575,
      "learning_rate": 0.00033785105615189,
      "loss": 1.6540666818618774,
      "step": 10240
    },
    {
      "epoch": 0.472957369174634,
      "eval_cos_loss": 0.40575925862952456,
      "eval_loss": 1.6192435720195508,
      "eval_mse_loss": 1.213484312301357,
      "flow/cos_sim": 0.5942407566118458,
      "flow/improvement_ratio": 0.9697231067914397,
      "flow/mag_ratio_mean": 0.5884986310244695,
      "flow/mag_ratio_std": 0.2199755361703433,
      "step": 10240
    },
    {
      "epoch": 0.472957369174634,
      "eval_cos_loss": 0.40575925862952456,
      "eval_loss": 1.6192435720195508,
      "eval_mse_loss": 1.213484312301357,
      "eval_runtime": 39.5214,
      "eval_samples_per_second": 708.3,
      "eval_steps_per_second": 11.083,
      "flow/cos_sim": 0.5942407566118458,
      "flow/improvement_ratio": 0.9697231067914397,
      "flow/mag_ratio_mean": 0.5884986310244695,
      "flow/mag_ratio_std": 0.2199755361703433,
      "step": 10240
    },
    {
      "epoch": 0.5202531060920974,
      "grad_norm": 0.35981205105781555,
      "learning_rate": 0.0002962123380271837,
      "loss": 1.6480820178985596,
      "step": 11264
    },
    {
      "epoch": 0.5202531060920974,
      "eval_cos_loss": 0.4020878194265714,
      "eval_loss": 1.6067346075898437,
      "eval_mse_loss": 1.2046467928581586,
      "flow/cos_sim": 0.5979121977880121,
      "flow/improvement_ratio": 0.9696458948257307,
      "flow/mag_ratio_mean": 0.5931163091365605,
      "flow/mag_ratio_std": 0.2266381677563332,
      "step": 11264
    },
    {
      "epoch": 0.5202531060920974,
      "eval_cos_loss": 0.4020878194265714,
      "eval_loss": 1.6067346075898437,
      "eval_mse_loss": 1.2046467928581586,
      "eval_runtime": 40.0623,
      "eval_samples_per_second": 698.736,
      "eval_steps_per_second": 10.933,
      "flow/cos_sim": 0.5979121977880121,
      "flow/improvement_ratio": 0.9696458948257307,
      "flow/mag_ratio_mean": 0.5931163091365605,
      "flow/mag_ratio_std": 0.2266381677563332,
      "step": 11264
    },
    {
      "epoch": 0.5675488430095608,
      "grad_norm": 0.5238872766494751,
      "learning_rate": 0.00025323382764865535,
      "loss": 1.6410928964614868,
      "step": 12288
    },
    {
      "epoch": 0.5675488430095608,
      "eval_cos_loss": 0.4016706557714776,
      "eval_loss": 1.6051203358118937,
      "eval_mse_loss": 1.2034496806527926,
      "flow/cos_sim": 0.5983293527337514,
      "flow/improvement_ratio": 0.9703199659308342,
      "flow/mag_ratio_mean": 0.5952222302351913,
      "flow/mag_ratio_std": 0.2251834258709324,
      "step": 12288
    },
    {
      "epoch": 0.5675488430095608,
      "eval_cos_loss": 0.4016706557714776,
      "eval_loss": 1.6051203358118937,
      "eval_mse_loss": 1.2034496806527926,
      "eval_runtime": 39.8137,
      "eval_samples_per_second": 703.099,
      "eval_steps_per_second": 11.001,
      "flow/cos_sim": 0.5983293527337514,
      "flow/improvement_ratio": 0.9703199659308342,
      "flow/mag_ratio_mean": 0.5952222302351913,
      "flow/mag_ratio_std": 0.2251834258709324,
      "step": 12288
    },
    {
      "epoch": 0.6148445799270241,
      "grad_norm": 0.26784732937812805,
      "learning_rate": 0.000210117059628806,
      "loss": 1.63731050491333,
      "step": 13312
    },
    {
      "epoch": 0.6148445799270241,
      "eval_cos_loss": 0.4014841335140951,
      "eval_loss": 1.604530786814755,
      "eval_mse_loss": 1.2030466533687016,
      "flow/cos_sim": 0.5985158796179785,
      "flow/improvement_ratio": 0.9693736963348302,
      "flow/mag_ratio_mean": 0.5919299127032224,
      "flow/mag_ratio_std": 0.22532628976727184,
      "step": 13312
    },
    {
      "epoch": 0.6148445799270241,
      "eval_cos_loss": 0.4014841335140951,
      "eval_loss": 1.604530786814755,
      "eval_mse_loss": 1.2030466533687016,
      "eval_runtime": 40.496,
      "eval_samples_per_second": 691.254,
      "eval_steps_per_second": 10.816,
      "flow/cos_sim": 0.5985158796179785,
      "flow/improvement_ratio": 0.9693736963348302,
      "flow/mag_ratio_mean": 0.5919299127032224,
      "flow/mag_ratio_std": 0.22532628976727184,
      "step": 13312
    },
    {
      "epoch": 0.6621403168444876,
      "grad_norm": 0.43763285875320435,
      "learning_rate": 0.0001682330091706446,
      "loss": 1.6302827596664429,
      "step": 14336
    },
    {
      "epoch": 0.6621403168444876,
      "eval_cos_loss": 0.3985591554614507,
      "eval_loss": 1.593907880184313,
      "eval_mse_loss": 1.1953487265599918,
      "flow/cos_sim": 0.6014408363054876,
      "flow/improvement_ratio": 0.9697505197568571,
      "flow/mag_ratio_mean": 0.5954256657868215,
      "flow/mag_ratio_std": 0.2259266352340511,
      "step": 14336
    },
    {
      "epoch": 0.6621403168444876,
      "eval_cos_loss": 0.3985591554614507,
      "eval_loss": 1.593907880184313,
      "eval_mse_loss": 1.1953487265599918,
      "eval_runtime": 40.0371,
      "eval_samples_per_second": 699.177,
      "eval_steps_per_second": 10.94,
      "flow/cos_sim": 0.6014408363054876,
      "flow/improvement_ratio": 0.9697505197568571,
      "flow/mag_ratio_mean": 0.5954256657868215,
      "flow/mag_ratio_std": 0.2259266352340511,
      "step": 14336
    },
    {
      "epoch": 0.709436053761951,
      "grad_norm": 0.341217041015625,
      "learning_rate": 0.0001287526134679914,
      "loss": 1.6316653490066528,
      "step": 15360
    },
    {
      "epoch": 0.709436053761951,
      "eval_cos_loss": 0.3979599736460812,
      "eval_loss": 1.5915418874183203,
      "eval_mse_loss": 1.1935819131598624,
      "flow/cos_sim": 0.6020400329539765,
      "flow/improvement_ratio": 0.9696243418950469,
      "flow/mag_ratio_mean": 0.595387307612319,
      "flow/mag_ratio_std": 0.22555491620819318,
      "step": 15360
    },
    {
      "epoch": 0.709436053761951,
      "eval_cos_loss": 0.3979599736460812,
      "eval_loss": 1.5915418874183203,
      "eval_mse_loss": 1.1935819131598624,
      "eval_runtime": 39.8928,
      "eval_samples_per_second": 701.706,
      "eval_steps_per_second": 10.979,
      "flow/cos_sim": 0.6020400329539765,
      "flow/improvement_ratio": 0.9696243418950469,
      "flow/mag_ratio_mean": 0.595387307612319,
      "flow/mag_ratio_std": 0.22555491620819318,
      "step": 15360
    },
    {
      "epoch": 0.7567317906794143,
      "grad_norm": 0.25786092877388,
      "learning_rate": 9.293122272447036e-05,
      "loss": 1.6261003017425537,
      "step": 16384
    },
    {
      "epoch": 0.7567317906794143,
      "eval_cos_loss": 0.3959788013947064,
      "eval_loss": 1.5846927321120485,
      "eval_mse_loss": 1.1887139288802124,
      "flow/cos_sim": 0.6040212045249329,
      "flow/improvement_ratio": 0.9709213373323554,
      "flow/mag_ratio_mean": 0.5975744849470652,
      "flow/mag_ratio_std": 0.2268577309912198,
      "step": 16384
    },
    {
      "epoch": 0.7567317906794143,
      "eval_cos_loss": 0.3959788013947064,
      "eval_loss": 1.5846927321120485,
      "eval_mse_loss": 1.1887139288802124,
      "eval_runtime": 40.2532,
      "eval_samples_per_second": 695.423,
      "eval_steps_per_second": 10.881,
      "flow/cos_sim": 0.6040212045249329,
      "flow/improvement_ratio": 0.9709213373323554,
      "flow/mag_ratio_mean": 0.5975744849470652,
      "flow/mag_ratio_std": 0.2268577309912198,
      "step": 16384
    },
    {
      "epoch": 0.8040275275968778,
      "grad_norm": 0.30163654685020447,
      "learning_rate": 6.177028255713927e-05,
      "loss": 1.6237930059432983,
      "step": 17408
    },
    {
      "epoch": 0.8040275275968778,
      "eval_cos_loss": 0.397802642967603,
      "eval_loss": 1.5906421003276354,
      "eval_mse_loss": 1.1928394578363253,
      "flow/cos_sim": 0.6021973686675502,
      "flow/improvement_ratio": 0.9696774086723589,
      "flow/mag_ratio_mean": 0.5974947977555941,
      "flow/mag_ratio_std": 0.22862732230120053,
      "step": 17408
    },
    {
      "epoch": 0.8040275275968778,
      "eval_cos_loss": 0.397802642967603,
      "eval_loss": 1.5906421003276354,
      "eval_mse_loss": 1.1928394578363253,
      "eval_runtime": 39.4241,
      "eval_samples_per_second": 710.047,
      "eval_steps_per_second": 11.11,
      "flow/cos_sim": 0.6021973686675502,
      "flow/improvement_ratio": 0.9696774086723589,
      "flow/mag_ratio_mean": 0.5974947977555941,
      "flow/mag_ratio_std": 0.22862732230120053,
      "step": 17408
    },
    {
      "epoch": 0.8513232645143411,
      "grad_norm": 0.3143230974674225,
      "learning_rate": 3.623868702551317e-05,
      "loss": 1.623117446899414,
      "step": 18432
    },
    {
      "epoch": 0.8513232645143411,
      "eval_cos_loss": 0.39859252644183973,
      "eval_loss": 1.5927672549469831,
      "eval_mse_loss": 1.1941747259875954,
      "flow/cos_sim": 0.6014074842407279,
      "flow/improvement_ratio": 0.9700041171078269,
      "flow/mag_ratio_mean": 0.5950140223655527,
      "flow/mag_ratio_std": 0.22678067565780796,
      "step": 18432
    },
    {
      "epoch": 0.8513232645143411,
      "eval_cos_loss": 0.39859252644183973,
      "eval_loss": 1.5927672549469831,
      "eval_mse_loss": 1.1941747259875954,
      "eval_runtime": 40.1556,
      "eval_samples_per_second": 697.113,
      "eval_steps_per_second": 10.908,
      "flow/cos_sim": 0.6014074842407279,
      "flow/improvement_ratio": 0.9700041171078269,
      "flow/mag_ratio_mean": 0.5950140223655527,
      "flow/mag_ratio_std": 0.22678067565780796,
      "step": 18432
    },
    {
      "epoch": 0.8986190014318045,
      "grad_norm": 0.23344755172729492,
      "learning_rate": 1.710000392483449e-05,
      "loss": 1.6220356225967407,
      "step": 19456
    },
    {
      "epoch": 0.8986190014318045,
      "eval_cos_loss": 0.3966924757717951,
      "eval_loss": 1.5871477426459255,
      "eval_mse_loss": 1.1904552660576284,
      "flow/cos_sim": 0.6033075423273322,
      "flow/improvement_ratio": 0.9706315692156962,
      "flow/mag_ratio_mean": 0.5972542343618663,
      "flow/mag_ratio_std": 0.22712956321293906,
      "step": 19456
    },
    {
      "epoch": 0.8986190014318045,
      "eval_cos_loss": 0.3966924757717951,
      "eval_loss": 1.5871477426459255,
      "eval_mse_loss": 1.1904552660576284,
      "eval_runtime": 39.2746,
      "eval_samples_per_second": 712.751,
      "eval_steps_per_second": 11.152,
      "flow/cos_sim": 0.6033075423273322,
      "flow/improvement_ratio": 0.9706315692156962,
      "flow/mag_ratio_mean": 0.5972542343618663,
      "flow/mag_ratio_std": 0.22712956321293906,
      "step": 19456
    }
  ],
  "logging_steps": 1024,
  "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
}