{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.287374362065919,
  "eval_steps": 1024,
  "global_step": 27648,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.010643494891330332,
      "grad_norm": 0.13342437148094177,
      "learning_rate": 0.0003330078125,
      "loss": 2.2998437881469727,
      "step": 1024
    },
    {
      "epoch": 0.010643494891330332,
      "eval_cos_loss": 0.5988449528813362,
      "eval_loss": 1.9600126259028912,
      "eval_mse_loss": 1.6605901420116425,
      "flow/cos_sim": 0.4011551085859537,
      "flow/improvement_ratio": 0.942170824855566,
      "flow/mag_ratio_mean": 0.37856073677539825,
      "flow/mag_ratio_std": 0.14085532305762172,
      "step": 1024
    },
    {
      "epoch": 0.010643494891330332,
      "eval_cos_loss": 0.5988449528813362,
      "eval_loss": 1.9600126259028912,
      "eval_mse_loss": 1.6605901420116425,
      "eval_runtime": 2.6584,
      "eval_samples_per_second": 752.329,
      "eval_steps_per_second": 12.037,
      "flow/cos_sim": 0.4011551085859537,
      "flow/improvement_ratio": 0.942170824855566,
      "flow/mag_ratio_mean": 0.37856073677539825,
      "flow/mag_ratio_std": 0.14085532305762172,
      "step": 1024
    },
    {
      "epoch": 0.021286989782660665,
      "grad_norm": 0.25054192543029785,
      "learning_rate": 0.0006663411458333333,
      "loss": 1.8492329120635986,
      "step": 2048
    },
    {
      "epoch": 0.021286989782660665,
      "eval_cos_loss": 0.5117531130090356,
      "eval_loss": 1.7429817728698254,
      "eval_mse_loss": 1.4871052131056786,
      "flow/cos_sim": 0.4882468534633517,
      "flow/improvement_ratio": 0.9563530795276165,
      "flow/mag_ratio_mean": 0.47669631242752075,
      "flow/mag_ratio_std": 0.17675806442275643,
      "step": 2048
    },
    {
      "epoch": 0.021286989782660665,
      "eval_cos_loss": 0.5117531130090356,
      "eval_loss": 1.7429817728698254,
      "eval_mse_loss": 1.4871052131056786,
      "eval_runtime": 2.511,
      "eval_samples_per_second": 796.509,
      "eval_steps_per_second": 12.744,
      "flow/cos_sim": 0.4882468534633517,
      "flow/improvement_ratio": 0.9563530795276165,
      "flow/mag_ratio_mean": 0.47669631242752075,
      "flow/mag_ratio_std": 0.17675806442275643,
      "step": 2048
    },
    {
      "epoch": 0.031930484673991,
      "grad_norm": 0.30941224098205566,
      "learning_rate": 0.0009996744791666667,
      "loss": 1.730944037437439,
      "step": 3072
    },
    {
      "epoch": 0.031930484673991,
      "eval_cos_loss": 0.4815286351367831,
      "eval_loss": 1.6586528308689594,
      "eval_mse_loss": 1.4178885221481323,
      "flow/cos_sim": 0.5184714393690228,
      "flow/improvement_ratio": 0.9605911839753389,
      "flow/mag_ratio_mean": 0.49818364903330803,
      "flow/mag_ratio_std": 0.1928270636126399,
      "step": 3072
    },
    {
      "epoch": 0.031930484673991,
      "eval_cos_loss": 0.4815286351367831,
      "eval_loss": 1.6586528308689594,
      "eval_mse_loss": 1.4178885221481323,
      "eval_runtime": 3.1033,
      "eval_samples_per_second": 644.485,
      "eval_steps_per_second": 10.312,
      "flow/cos_sim": 0.5184714393690228,
      "flow/improvement_ratio": 0.9605911839753389,
      "flow/mag_ratio_mean": 0.49818364903330803,
      "flow/mag_ratio_std": 0.1928270636126399,
      "step": 3072
    },
    {
      "epoch": 0.04257397956532133,
      "grad_norm": 0.22964967787265778,
      "learning_rate": 0.0009997023516784352,
      "loss": 1.6850833892822266,
      "step": 4096
    },
    {
      "epoch": 0.04257397956532133,
      "eval_cos_loss": 0.476364528760314,
      "eval_loss": 1.6391540355980396,
      "eval_mse_loss": 1.4009717665612698,
      "flow/cos_sim": 0.5236354488879442,
      "flow/improvement_ratio": 0.9618693646043539,
      "flow/mag_ratio_mean": 0.5105963433161378,
      "flow/mag_ratio_std": 0.20592432795092463,
      "step": 4096
    },
    {
      "epoch": 0.04257397956532133,
      "eval_cos_loss": 0.476364528760314,
      "eval_loss": 1.6391540355980396,
      "eval_mse_loss": 1.4009717665612698,
      "eval_runtime": 2.5129,
      "eval_samples_per_second": 795.895,
      "eval_steps_per_second": 12.734,
      "flow/cos_sim": 0.5236354488879442,
      "flow/improvement_ratio": 0.9618693646043539,
      "flow/mag_ratio_mean": 0.5105963433161378,
      "flow/mag_ratio_std": 0.20592432795092463,
      "step": 4096
    },
    {
      "epoch": 0.05321747445665166,
      "grad_norm": 0.2645546495914459,
      "learning_rate": 0.0009988085977910004,
      "loss": 1.6617510318756104,
      "step": 5120
    },
    {
      "epoch": 0.05321747445665166,
      "eval_cos_loss": 0.4789119102060795,
      "eval_loss": 1.645260013639927,
      "eval_mse_loss": 1.405804067850113,
      "flow/cos_sim": 0.5210880534723401,
      "flow/improvement_ratio": 0.9588682930916548,
      "flow/mag_ratio_mean": 0.504288200289011,
      "flow/mag_ratio_std": 0.20718340016901493,
      "step": 5120
    },
    {
      "epoch": 0.05321747445665166,
      "eval_cos_loss": 0.4789119102060795,
      "eval_loss": 1.645260013639927,
      "eval_mse_loss": 1.405804067850113,
      "eval_runtime": 3.0976,
      "eval_samples_per_second": 645.654,
      "eval_steps_per_second": 10.33,
      "flow/cos_sim": 0.5210880534723401,
      "flow/improvement_ratio": 0.9588682930916548,
      "flow/mag_ratio_mean": 0.504288200289011,
      "flow/mag_ratio_std": 0.20718340016901493,
      "step": 5120
    },
    {
      "epoch": 0.063860969347982,
      "grad_norm": 0.2762889862060547,
      "learning_rate": 0.0009973198042317873,
      "loss": 1.645796775817871,
      "step": 6144
    },
    {
      "epoch": 0.063860969347982,
      "eval_cos_loss": 0.4598818449303508,
      "eval_loss": 1.5948525853455067,
      "eval_mse_loss": 1.3649116680026054,
      "flow/cos_sim": 0.5401182025671005,
      "flow/improvement_ratio": 0.9647715575993061,
      "flow/mag_ratio_mean": 0.5178880272433162,
      "flow/mag_ratio_std": 0.21153279254212976,
      "step": 6144
    },
    {
      "epoch": 0.063860969347982,
      "eval_cos_loss": 0.4598818449303508,
      "eval_loss": 1.5948525853455067,
      "eval_mse_loss": 1.3649116680026054,
      "eval_runtime": 3.0831,
      "eval_samples_per_second": 648.695,
      "eval_steps_per_second": 10.379,
      "flow/cos_sim": 0.5401182025671005,
      "flow/improvement_ratio": 0.9647715575993061,
      "flow/mag_ratio_mean": 0.5178880272433162,
      "flow/mag_ratio_std": 0.21153279254212976,
      "step": 6144
    },
    {
      "epoch": 0.07450446423931233,
      "grad_norm": 0.17679959535598755,
      "learning_rate": 0.0009952377470151526,
      "loss": 1.6353809833526611,
      "step": 7168
    },
    {
      "epoch": 0.07450446423931233,
      "eval_cos_loss": 0.4634226718917489,
      "eval_loss": 1.6022505089640617,
      "eval_mse_loss": 1.3705391697585583,
      "flow/cos_sim": 0.5365773290395737,
      "flow/improvement_ratio": 0.9635819494724274,
      "flow/mag_ratio_mean": 0.5194354858249426,
      "flow/mag_ratio_std": 0.21515046246349812,
      "step": 7168
    },
    {
      "epoch": 0.07450446423931233,
      "eval_cos_loss": 0.4634226718917489,
      "eval_loss": 1.6022505089640617,
      "eval_mse_loss": 1.3705391697585583,
      "eval_runtime": 2.8419,
      "eval_samples_per_second": 703.759,
      "eval_steps_per_second": 11.26,
      "flow/cos_sim": 0.5365773290395737,
      "flow/improvement_ratio": 0.9635819494724274,
      "flow/mag_ratio_mean": 0.5194354858249426,
      "flow/mag_ratio_std": 0.21515046246349812,
      "step": 7168
    },
    {
      "epoch": 0.08514795913064266,
      "grad_norm": 0.14975515007972717,
      "learning_rate": 0.000992564909872628,
      "loss": 1.6262034177780151,
      "step": 8192
    },
    {
      "epoch": 0.08514795913064266,
      "eval_cos_loss": 0.45912545546889305,
      "eval_loss": 1.5890175811946392,
      "eval_mse_loss": 1.3594548553228378,
      "flow/cos_sim": 0.5408745482563972,
      "flow/improvement_ratio": 0.9590303134173155,
      "flow/mag_ratio_mean": 0.5143361240625381,
      "flow/mag_ratio_std": 0.21537457825616002,
      "step": 8192
    },
    {
      "epoch": 0.08514795913064266,
      "eval_cos_loss": 0.45912545546889305,
      "eval_loss": 1.5890175811946392,
      "eval_mse_loss": 1.3594548553228378,
      "eval_runtime": 2.9302,
      "eval_samples_per_second": 682.537,
      "eval_steps_per_second": 10.921,
      "flow/cos_sim": 0.5408745482563972,
      "flow/improvement_ratio": 0.9590303134173155,
      "flow/mag_ratio_mean": 0.5143361240625381,
      "flow/mag_ratio_std": 0.21537457825616002,
      "step": 8192
    },
    {
      "epoch": 0.09579145402197299,
      "grad_norm": 0.19106586277484894,
      "learning_rate": 0.000989307950724573,
      "loss": 1.6214015483856201,
      "step": 9216
    },
    {
      "epoch": 0.09579145402197299,
      "eval_cos_loss": 0.4567577252164483,
      "eval_loss": 1.5844898335635662,
      "eval_mse_loss": 1.356110967695713,
      "flow/cos_sim": 0.5432424321770668,
      "flow/improvement_ratio": 0.9650511220097542,
      "flow/mag_ratio_mean": 0.5244949720799923,
      "flow/mag_ratio_std": 0.21130397450178862,
      "step": 9216
    },
    {
      "epoch": 0.09579145402197299,
      "eval_cos_loss": 0.4567577252164483,
      "eval_loss": 1.5844898335635662,
      "eval_mse_loss": 1.356110967695713,
      "eval_runtime": 3.0508,
      "eval_samples_per_second": 655.562,
      "eval_steps_per_second": 10.489,
      "flow/cos_sim": 0.5432424321770668,
      "flow/improvement_ratio": 0.9650511220097542,
      "flow/mag_ratio_mean": 0.5244949720799923,
      "flow/mag_ratio_std": 0.21130397450178862,
      "step": 9216
    },
    {
      "epoch": 0.10643494891330332,
      "grad_norm": 0.22245089709758759,
      "learning_rate": 0.000985464388035817,
      "loss": 1.6132733821868896,
      "step": 10240
    },
    {
      "epoch": 0.10643494891330332,
      "eval_cos_loss": 0.4598613306879997,
      "eval_loss": 1.589576181024313,
      "eval_mse_loss": 1.359645515680313,
      "flow/cos_sim": 0.5401386898010969,
      "flow/improvement_ratio": 0.9610863700509071,
      "flow/mag_ratio_mean": 0.5160716716200113,
      "flow/mag_ratio_std": 0.21545762522146106,
      "step": 10240
    },
    {
      "epoch": 0.10643494891330332,
      "eval_cos_loss": 0.4598613306879997,
      "eval_loss": 1.589576181024313,
      "eval_mse_loss": 1.359645515680313,
      "eval_runtime": 3.1847,
      "eval_samples_per_second": 627.993,
      "eval_steps_per_second": 10.048,
      "flow/cos_sim": 0.5401386898010969,
      "flow/improvement_ratio": 0.9610863700509071,
      "flow/mag_ratio_mean": 0.5160716716200113,
      "flow/mag_ratio_std": 0.21545762522146106,
      "step": 10240
    },
    {
      "epoch": 0.11707844380463366,
      "grad_norm": 0.1567550003528595,
      "learning_rate": 0.0009810417042745768,
      "loss": 1.6070518493652344,
      "step": 11264
    },
    {
      "epoch": 0.11707844380463366,
      "eval_cos_loss": 0.4550258554518223,
      "eval_loss": 1.577816877514124,
      "eval_mse_loss": 1.350303951650858,
      "flow/cos_sim": 0.5449741557240486,
      "flow/improvement_ratio": 0.9648044053465128,
      "flow/mag_ratio_mean": 0.5290831215679646,
      "flow/mag_ratio_std": 0.21279342425987124,
      "step": 11264
    },
    {
      "epoch": 0.11707844380463366,
      "eval_cos_loss": 0.4550258554518223,
      "eval_loss": 1.577816877514124,
      "eval_mse_loss": 1.350303951650858,
      "eval_runtime": 2.6352,
      "eval_samples_per_second": 758.945,
      "eval_steps_per_second": 12.143,
      "flow/cos_sim": 0.5449741557240486,
      "flow/improvement_ratio": 0.9648044053465128,
      "flow/mag_ratio_mean": 0.5290831215679646,
      "flow/mag_ratio_std": 0.21279342425987124,
      "step": 11264
    },
    {
      "epoch": 0.127721938695964,
      "grad_norm": 0.19316641986370087,
      "learning_rate": 0.0009760451753569162,
      "loss": 1.6028146743774414,
      "step": 12288
    },
    {
      "epoch": 0.127721938695964,
      "eval_cos_loss": 0.45873888209462166,
      "eval_loss": 1.5860362015664577,
      "eval_mse_loss": 1.3566667586565018,
      "flow/cos_sim": 0.5412612538784742,
      "flow/improvement_ratio": 0.9610528890043497,
      "flow/mag_ratio_mean": 0.5225661229342222,
      "flow/mag_ratio_std": 0.2149493475444615,
      "step": 12288
    },
    {
      "epoch": 0.127721938695964,
      "eval_cos_loss": 0.45873888209462166,
      "eval_loss": 1.5860362015664577,
      "eval_mse_loss": 1.3566667586565018,
      "eval_runtime": 2.6091,
      "eval_samples_per_second": 766.55,
      "eval_steps_per_second": 12.265,
      "flow/cos_sim": 0.5412612538784742,
      "flow/improvement_ratio": 0.9610528890043497,
      "flow/mag_ratio_mean": 0.5225661229342222,
      "flow/mag_ratio_std": 0.2149493475444615,
      "step": 12288
    },
    {
      "epoch": 0.13836543358729433,
      "grad_norm": 0.17066629230976105,
      "learning_rate": 0.000970486470662755,
      "loss": 1.5989067554473877,
      "step": 13312
    },
    {
      "epoch": 0.13836543358729433,
      "eval_cos_loss": 0.4526587063446641,
      "eval_loss": 1.5703520886600018,
      "eval_mse_loss": 1.3440227322280407,
      "flow/cos_sim": 0.547341376543045,
      "flow/improvement_ratio": 0.9634687285870314,
      "flow/mag_ratio_mean": 0.5251006819307804,
      "flow/mag_ratio_std": 0.2169443154707551,
      "step": 13312
    },
    {
      "epoch": 0.13836543358729433,
      "eval_cos_loss": 0.4526587063446641,
      "eval_loss": 1.5703520886600018,
      "eval_mse_loss": 1.3440227322280407,
      "eval_runtime": 2.6502,
      "eval_samples_per_second": 754.652,
      "eval_steps_per_second": 12.074,
      "flow/cos_sim": 0.547341376543045,
      "flow/improvement_ratio": 0.9634687285870314,
      "flow/mag_ratio_mean": 0.5251006819307804,
      "flow/mag_ratio_std": 0.2169443154707551,
      "step": 13312
    },
    {
      "epoch": 0.14900892847862465,
      "grad_norm": 0.19086262583732605,
      "learning_rate": 0.0009643613549160033,
      "loss": 1.5941526889801025,
      "step": 14336
    },
    {
      "epoch": 0.14900892847862465,
      "eval_cos_loss": 0.45674111880362034,
      "eval_loss": 1.5803881026804447,
      "eval_mse_loss": 1.3520175516605377,
      "flow/cos_sim": 0.5432589612901211,
      "flow/improvement_ratio": 0.9569191709160805,
      "flow/mag_ratio_mean": 0.5241195531561971,
      "flow/mag_ratio_std": 0.2207528604194522,
      "step": 14336
    },
    {
      "epoch": 0.14900892847862465,
      "eval_cos_loss": 0.45674111880362034,
      "eval_loss": 1.5803881026804447,
      "eval_mse_loss": 1.3520175516605377,
      "eval_runtime": 2.84,
      "eval_samples_per_second": 704.229,
      "eval_steps_per_second": 11.268,
      "flow/cos_sim": 0.5432589612901211,
      "flow/improvement_ratio": 0.9569191709160805,
      "flow/mag_ratio_mean": 0.5241195531561971,
      "flow/mag_ratio_std": 0.2207528604194522,
      "step": 14336
    },
    {
      "epoch": 0.159652423369955,
      "grad_norm": 0.20660291612148285,
      "learning_rate": 0.0009576890825691249,
      "loss": 1.5903245210647583,
      "step": 15360
    },
    {
      "epoch": 0.159652423369955,
      "eval_cos_loss": 0.4470532648265362,
      "eval_loss": 1.5533855073153973,
      "eval_mse_loss": 1.329858873039484,
      "flow/cos_sim": 0.5529466420412064,
      "flow/improvement_ratio": 0.9680595081299543,
      "flow/mag_ratio_mean": 0.5352848172187805,
      "flow/mag_ratio_std": 0.22097993176430464,
      "step": 15360
    },
    {
      "epoch": 0.159652423369955,
      "eval_cos_loss": 0.4470532648265362,
      "eval_loss": 1.5533855073153973,
      "eval_mse_loss": 1.329858873039484,
      "eval_runtime": 2.6578,
      "eval_samples_per_second": 752.494,
      "eval_steps_per_second": 12.04,
      "flow/cos_sim": 0.5529466420412064,
      "flow/improvement_ratio": 0.9680595081299543,
      "flow/mag_ratio_mean": 0.5352848172187805,
      "flow/mag_ratio_std": 0.22097993176430464,
      "step": 15360
    },
    {
      "epoch": 0.17029591826128532,
      "grad_norm": 0.23885692656040192,
      "learning_rate": 0.0009504645698990064,
      "loss": 1.589218020439148,
      "step": 16384
    },
    {
      "epoch": 0.17029591826128532,
      "eval_cos_loss": 0.44447089545428753,
      "eval_loss": 1.5484142042696476,
      "eval_mse_loss": 1.3261787556111813,
      "flow/cos_sim": 0.5555290877819061,
      "flow/improvement_ratio": 0.9635521955788136,
      "flow/mag_ratio_mean": 0.5299641713500023,
      "flow/mag_ratio_std": 0.215805409476161,
      "step": 16384
    },
    {
      "epoch": 0.17029591826128532,
      "eval_cos_loss": 0.44447089545428753,
      "eval_loss": 1.5484142042696476,
      "eval_mse_loss": 1.3261787556111813,
      "eval_runtime": 2.6371,
      "eval_samples_per_second": 758.402,
      "eval_steps_per_second": 12.134,
      "flow/cos_sim": 0.5555290877819061,
      "flow/improvement_ratio": 0.9635521955788136,
      "flow/mag_ratio_mean": 0.5299641713500023,
      "flow/mag_ratio_std": 0.215805409476161,
      "step": 16384
    },
    {
      "epoch": 0.18093941315261566,
      "grad_norm": 0.2062983363866806,
      "learning_rate": 0.0009427105273394636,
      "loss": 1.585401177406311,
      "step": 17408
    },
    {
      "epoch": 0.18093941315261566,
      "eval_cos_loss": 0.45094432309269905,
      "eval_loss": 1.5633347816765308,
      "eval_mse_loss": 1.3378626182675362,
      "flow/cos_sim": 0.5490557141602039,
      "flow/improvement_ratio": 0.9632246606051922,
      "flow/mag_ratio_mean": 0.5234426287934184,
      "flow/mag_ratio_std": 0.22044725203886628,
      "step": 17408
    },
    {
      "epoch": 0.18093941315261566,
      "eval_cos_loss": 0.45094432309269905,
      "eval_loss": 1.5633347816765308,
      "eval_mse_loss": 1.3378626182675362,
      "eval_runtime": 2.5802,
      "eval_samples_per_second": 775.125,
      "eval_steps_per_second": 12.402,
      "flow/cos_sim": 0.5490557141602039,
      "flow/improvement_ratio": 0.9632246606051922,
      "flow/mag_ratio_mean": 0.5234426287934184,
      "flow/mag_ratio_std": 0.22044725203886628,
      "step": 17408
    },
    {
      "epoch": 0.19158290804394598,
      "grad_norm": 0.15858766436576843,
      "learning_rate": 0.0009344210469473947,
      "loss": 1.5826770067214966,
      "step": 18432
    },
    {
      "epoch": 0.19158290804394598,
      "eval_cos_loss": 0.44898632261902094,
      "eval_loss": 1.5564597770571709,
      "eval_mse_loss": 1.331966608762741,
      "flow/cos_sim": 0.5510137844830751,
      "flow/improvement_ratio": 0.9625816307961941,
      "flow/mag_ratio_mean": 0.5298811597749591,
      "flow/mag_ratio_std": 0.22253544814884663,
      "step": 18432
    },
    {
      "epoch": 0.19158290804394598,
      "eval_cos_loss": 0.44898632261902094,
      "eval_loss": 1.5564597770571709,
      "eval_mse_loss": 1.331966608762741,
      "eval_runtime": 2.5531,
      "eval_samples_per_second": 783.347,
      "eval_steps_per_second": 12.534,
      "flow/cos_sim": 0.5510137844830751,
      "flow/improvement_ratio": 0.9625816307961941,
      "flow/mag_ratio_mean": 0.5298811597749591,
      "flow/mag_ratio_std": 0.22253544814884663,
      "step": 18432
    },
    {
      "epoch": 0.20222640293527633,
      "grad_norm": 0.2525703012943268,
      "learning_rate": 0.0009256133361993658,
      "loss": 1.5798900127410889,
      "step": 19456
    },
    {
      "epoch": 0.20222640293527633,
      "eval_cos_loss": 0.45141084399074316,
      "eval_loss": 1.567859135568142,
      "eval_mse_loss": 1.3421537093818188,
      "flow/cos_sim": 0.5485891196876764,
      "flow/improvement_ratio": 0.96523248963058,
      "flow/mag_ratio_mean": 0.5181732634082437,
      "flow/mag_ratio_std": 0.22030179109424353,
      "step": 19456
    },
    {
      "epoch": 0.20222640293527633,
      "eval_cos_loss": 0.45141084399074316,
      "eval_loss": 1.567859135568142,
      "eval_mse_loss": 1.3421537093818188,
      "eval_runtime": 2.5625,
      "eval_samples_per_second": 780.491,
      "eval_steps_per_second": 12.488,
      "flow/cos_sim": 0.5485891196876764,
      "flow/improvement_ratio": 0.96523248963058,
      "flow/mag_ratio_mean": 0.5181732634082437,
      "flow/mag_ratio_std": 0.22030179109424353,
      "step": 19456
    },
    {
      "epoch": 0.21286989782660665,
      "grad_norm": 0.18312996625900269,
      "learning_rate": 0.0009163072432159066,
      "loss": 1.579535961151123,
      "step": 20480
    },
    {
      "epoch": 0.21286989782660665,
      "eval_cos_loss": 0.45327545143663883,
      "eval_loss": 1.5679056644439697,
      "eval_mse_loss": 1.3412679433822632,
      "flow/cos_sim": 0.5467245355248451,
      "flow/improvement_ratio": 0.9616729654371738,
      "flow/mag_ratio_mean": 0.5273217614740133,
      "flow/mag_ratio_std": 0.2235504975542426,
      "step": 20480
    },
    {
      "epoch": 0.21286989782660665,
      "eval_cos_loss": 0.45327545143663883,
      "eval_loss": 1.5679056644439697,
      "eval_mse_loss": 1.3412679433822632,
      "eval_runtime": 2.5697,
      "eval_samples_per_second": 778.293,
      "eval_steps_per_second": 12.453,
      "flow/cos_sim": 0.5467245355248451,
      "flow/improvement_ratio": 0.9616729654371738,
      "flow/mag_ratio_mean": 0.5273217614740133,
      "flow/mag_ratio_std": 0.2235504975542426,
      "step": 20480
    },
    {
      "epoch": 0.223513392717937,
      "grad_norm": 0.21262691915035248,
      "learning_rate": 0.0009064956775190607,
      "loss": 1.577104926109314,
      "step": 21504
    },
    {
      "epoch": 0.223513392717937,
      "eval_cos_loss": 0.4483450762927532,
      "eval_loss": 1.553330171853304,
      "eval_mse_loss": 1.329157643020153,
      "flow/cos_sim": 0.5516549795866013,
      "flow/improvement_ratio": 0.9627660047262907,
      "flow/mag_ratio_mean": 0.5387043142691255,
      "flow/mag_ratio_std": 0.2251730626448989,
      "step": 21504
    },
    {
      "epoch": 0.223513392717937,
      "eval_cos_loss": 0.4483450762927532,
      "eval_loss": 1.553330171853304,
      "eval_mse_loss": 1.329157643020153,
      "eval_runtime": 2.5071,
      "eval_samples_per_second": 797.724,
      "eval_steps_per_second": 12.764,
      "flow/cos_sim": 0.5516549795866013,
      "flow/improvement_ratio": 0.9627660047262907,
      "flow/mag_ratio_mean": 0.5387043142691255,
      "flow/mag_ratio_std": 0.2251730626448989,
      "step": 21504
    },
    {
      "epoch": 0.2341568876092673,
      "grad_norm": 0.17988671362400055,
      "learning_rate": 0.0008961991942494195,
      "loss": 1.574266791343689,
      "step": 22528
    },
    {
      "epoch": 0.2341568876092673,
      "eval_cos_loss": 0.44411917496472597,
      "eval_loss": 1.543789055198431,
      "eval_mse_loss": 1.3217294700443745,
      "flow/cos_sim": 0.5558808352798223,
      "flow/improvement_ratio": 0.9671246875077486,
      "flow/mag_ratio_mean": 0.5348946927115321,
      "flow/mag_ratio_std": 0.22368196118623018,
      "step": 22528
    },
    {
      "epoch": 0.2341568876092673,
      "eval_cos_loss": 0.44411917496472597,
      "eval_loss": 1.543789055198431,
      "eval_mse_loss": 1.3217294700443745,
      "eval_runtime": 2.5967,
      "eval_samples_per_second": 770.215,
      "eval_steps_per_second": 12.323,
      "flow/cos_sim": 0.5558808352798223,
      "flow/improvement_ratio": 0.9671246875077486,
      "flow/mag_ratio_mean": 0.5348946927115321,
      "flow/mag_ratio_std": 0.22368196118623018,
      "step": 22528
    },
    {
      "epoch": 0.24480038250059766,
      "grad_norm": 0.22547593712806702,
      "learning_rate": 0.0008854408194461756,
      "loss": 1.5733323097229004,
      "step": 23552
    },
    {
      "epoch": 0.24480038250059766,
      "eval_cos_loss": 0.44172694999724627,
      "eval_loss": 1.5377833917737007,
      "eval_mse_loss": 1.316919919103384,
      "flow/cos_sim": 0.5582730043679476,
      "flow/improvement_ratio": 0.9642701335251331,
      "flow/mag_ratio_mean": 0.5346939843147993,
      "flow/mag_ratio_std": 0.22327208751812577,
      "step": 23552
    },
    {
      "epoch": 0.24480038250059766,
      "eval_cos_loss": 0.44172694999724627,
      "eval_loss": 1.5377833917737007,
      "eval_mse_loss": 1.316919919103384,
      "eval_runtime": 3.1395,
      "eval_samples_per_second": 637.04,
      "eval_steps_per_second": 10.193,
      "flow/cos_sim": 0.5582730043679476,
      "flow/improvement_ratio": 0.9642701335251331,
      "flow/mag_ratio_mean": 0.5346939843147993,
      "flow/mag_ratio_std": 0.22327208751812577,
      "step": 23552
    },
    {
      "epoch": 0.255443877391928,
      "grad_norm": 0.2300369143486023,
      "learning_rate": 0.0008742123561119935,
      "loss": 1.569944143295288,
      "step": 24576
    },
    {
      "epoch": 0.255443877391928,
      "eval_cos_loss": 0.447942478582263,
      "eval_loss": 1.553868442773819,
      "eval_mse_loss": 1.3298972100019455,
      "flow/cos_sim": 0.5520575055852532,
      "flow/improvement_ratio": 0.9638102632015944,
      "flow/mag_ratio_mean": 0.5306164929643273,
      "flow/mag_ratio_std": 0.22182104969397187,
      "step": 24576
    },
    {
      "epoch": 0.255443877391928,
      "eval_cos_loss": 0.447942478582263,
      "eval_loss": 1.553868442773819,
      "eval_mse_loss": 1.3298972100019455,
      "eval_runtime": 2.5886,
      "eval_samples_per_second": 772.612,
      "eval_steps_per_second": 12.362,
      "flow/cos_sim": 0.5520575055852532,
      "flow/improvement_ratio": 0.9638102632015944,
      "flow/mag_ratio_mean": 0.5306164929643273,
      "flow/mag_ratio_std": 0.22182104969397187,
      "step": 24576
    },
    {
      "epoch": 0.26608737228325835,
      "grad_norm": 0.2177908569574356,
      "learning_rate": 0.0008625491011983832,
      "loss": 1.5683772563934326,
      "step": 25600
    },
    {
      "epoch": 0.26608737228325835,
      "eval_cos_loss": 0.45118876080960035,
      "eval_loss": 1.5609249621629715,
      "eval_mse_loss": 1.3353305757045746,
      "flow/cos_sim": 0.5488111022859812,
      "flow/improvement_ratio": 0.9652206618338823,
      "flow/mag_ratio_mean": 0.5250881398096681,
      "flow/mag_ratio_std": 0.22340481635183096,
      "step": 25600
    },
    {
      "epoch": 0.26608737228325835,
      "eval_cos_loss": 0.45118876080960035,
      "eval_loss": 1.5609249621629715,
      "eval_mse_loss": 1.3353305757045746,
      "eval_runtime": 2.5832,
      "eval_samples_per_second": 774.241,
      "eval_steps_per_second": 12.388,
      "flow/cos_sim": 0.5488111022859812,
      "flow/improvement_ratio": 0.9652206618338823,
      "flow/mag_ratio_mean": 0.5250881398096681,
      "flow/mag_ratio_std": 0.22340481635183096,
      "step": 25600
    },
    {
      "epoch": 0.27673086717458867,
      "grad_norm": 0.13252000510692596,
      "learning_rate": 0.0008504421682637403,
      "loss": 1.5673582553863525,
      "step": 26624
    },
    {
      "epoch": 0.27673086717458867,
      "eval_cos_loss": 0.44563145097345114,
      "eval_loss": 1.5506689585745335,
      "eval_mse_loss": 1.3278532326221466,
      "flow/cos_sim": 0.5543686226010323,
      "flow/improvement_ratio": 0.9666622839868069,
      "flow/mag_ratio_mean": 0.5269411941990256,
      "flow/mag_ratio_std": 0.21877468656748533,
      "step": 26624
    },
    {
      "epoch": 0.27673086717458867,
      "eval_cos_loss": 0.44563145097345114,
      "eval_loss": 1.5506689585745335,
      "eval_mse_loss": 1.3278532326221466,
      "eval_runtime": 2.617,
      "eval_samples_per_second": 764.244,
      "eval_steps_per_second": 12.228,
      "flow/cos_sim": 0.5543686226010323,
      "flow/improvement_ratio": 0.9666622839868069,
      "flow/mag_ratio_mean": 0.5269411941990256,
      "flow/mag_ratio_std": 0.21877468656748533,
      "step": 26624
    },
    {
      "epoch": 0.287374362065919,
      "grad_norm": 0.2598721981048584,
      "learning_rate": 0.0008379296157504366,
      "loss": 1.564971923828125,
      "step": 27648
    },
    {
      "epoch": 0.287374362065919,
      "eval_cos_loss": 0.4419550793245435,
      "eval_loss": 1.5396056547760963,
      "eval_mse_loss": 1.3186281062662601,
      "flow/cos_sim": 0.5580449867993593,
      "flow/improvement_ratio": 0.9673310127109289,
      "flow/mag_ratio_mean": 0.5319117670878768,
      "flow/mag_ratio_std": 0.22065124148502946,
      "step": 27648
    },
    {
      "epoch": 0.287374362065919,
      "eval_cos_loss": 0.4419550793245435,
      "eval_loss": 1.5396056547760963,
      "eval_mse_loss": 1.3186281062662601,
      "eval_runtime": 3.0156,
      "eval_samples_per_second": 663.219,
      "eval_steps_per_second": 10.612,
      "flow/cos_sim": 0.5580449867993593,
      "flow/improvement_ratio": 0.9673310127109289,
      "flow/mag_ratio_mean": 0.5319117670878768,
      "flow/mag_ratio_std": 0.22065124148502946,
      "step": 27648
    }
  ],
  "logging_steps": 1024,
  "max_steps": 96209,
  "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
}