{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.4256616322571706,
  "eval_steps": 1024,
  "global_step": 9216,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.011823934229365849,
      "grad_norm": 0.07098234444856644,
      "learning_rate": 0.000498046875,
      "loss": 0.20588445663452148,
      "step": 256
    },
    {
      "epoch": 0.023647868458731697,
      "grad_norm": 0.22737464308738708,
      "learning_rate": 0.000998046875,
      "loss": 0.17480437457561493,
      "step": 512
    },
    {
      "epoch": 0.03547180268809755,
      "grad_norm": 0.12320689857006073,
      "learning_rate": 0.000999640996023194,
      "loss": 0.15027859807014465,
      "step": 768
    },
    {
      "epoch": 0.047295736917463395,
      "grad_norm": 0.20722734928131104,
      "learning_rate": 0.0009985588674043958,
      "loss": 0.1415732204914093,
      "step": 1024
    },
    {
      "epoch": 0.047295736917463395,
      "eval_cos_loss": 0.04582402333362053,
      "eval_loss": 0.09126683337333268,
      "eval_mse_loss": 0.09126683337333268,
      "flow/cos_sim": 0.9541759895135279,
      "flow/improvement_ratio": 0.9938432111587698,
      "flow/mag_ratio_mean": 0.8994761448200435,
      "flow/mag_ratio_std": 0.11755070905094822,
      "step": 1024
    },
    {
      "epoch": 0.047295736917463395,
      "eval_cos_loss": 0.04582402333362053,
      "eval_loss": 0.09126683337333268,
      "eval_mse_loss": 0.09126683337333268,
      "eval_runtime": 37.4141,
      "eval_samples_per_second": 748.193,
      "eval_steps_per_second": 11.707,
      "flow/cos_sim": 0.9541759895135279,
      "flow/improvement_ratio": 0.9938432111587698,
      "flow/mag_ratio_mean": 0.8994761448200435,
      "flow/mag_ratio_std": 0.11755070905094822,
      "step": 1024
    },
    {
      "epoch": 0.05911967114682925,
      "grad_norm": 0.12740713357925415,
      "learning_rate": 0.0009967551747861387,
      "loss": 0.13867560029029846,
      "step": 1280
    },
    {
      "epoch": 0.0709436053761951,
      "grad_norm": 0.49607324600219727,
      "learning_rate": 0.000994232528651847,
      "loss": 0.13481874763965607,
      "step": 1536
    },
    {
      "epoch": 0.08276753960556095,
      "grad_norm": 0.26182159781455994,
      "learning_rate": 0.0009909945800260092,
      "loss": 0.13546937704086304,
      "step": 1792
    },
    {
      "epoch": 0.09459147383492679,
      "grad_norm": 0.12674923241138458,
      "learning_rate": 0.0009870460151900522,
      "loss": 0.13297873735427856,
      "step": 2048
    },
    {
      "epoch": 0.09459147383492679,
      "eval_cos_loss": 0.04454593354470381,
      "eval_loss": 0.08455971572292995,
      "eval_mse_loss": 0.08455971572292995,
      "flow/cos_sim": 0.9554540837464267,
      "flow/improvement_ratio": 0.9942142926666835,
      "flow/mag_ratio_mean": 0.9197173737771979,
      "flow/mag_ratio_std": 0.1105972074448518,
      "step": 2048
    },
    {
      "epoch": 0.09459147383492679,
      "eval_cos_loss": 0.04454593354470381,
      "eval_loss": 0.08455971572292995,
      "eval_mse_loss": 0.08455971572292995,
      "eval_runtime": 36.8159,
      "eval_samples_per_second": 760.352,
      "eval_steps_per_second": 11.897,
      "flow/cos_sim": 0.9554540837464267,
      "flow/improvement_ratio": 0.9942142926666835,
      "flow/mag_ratio_mean": 0.9197173737771979,
      "flow/mag_ratio_std": 0.1105972074448518,
      "step": 2048
    },
    {
      "epoch": 0.10641540806429264,
      "grad_norm": 0.2195156067609787,
      "learning_rate": 0.0009823925488998885,
      "loss": 0.1306387186050415,
      "step": 2304
    },
    {
      "epoch": 0.1182393422936585,
      "grad_norm": 0.13101470470428467,
      "learning_rate": 0.0009770409161149525,
      "loss": 0.12852290272712708,
      "step": 2560
    },
    {
      "epoch": 0.13006327652302435,
      "grad_norm": 0.1463838666677475,
      "learning_rate": 0.0009709988622506973,
      "loss": 0.1289902925491333,
      "step": 2816
    },
    {
      "epoch": 0.1418872107523902,
      "grad_norm": 0.1557082235813141,
      "learning_rate": 0.000964275131968659,
      "loss": 0.12751106917858124,
      "step": 3072
    },
    {
      "epoch": 0.1418872107523902,
      "eval_cos_loss": 0.04429665551474105,
      "eval_loss": 0.08277356523391047,
      "eval_mse_loss": 0.08277356523391047,
      "flow/cos_sim": 0.9557033510512958,
      "flow/improvement_ratio": 0.9939037547808259,
      "flow/mag_ratio_mean": 0.92809867015168,
      "flow/mag_ratio_std": 0.10807815949394278,
      "step": 3072
    },
    {
      "epoch": 0.1418872107523902,
      "eval_cos_loss": 0.04429665551474105,
      "eval_loss": 0.08277356523391047,
      "eval_mse_loss": 0.08277356523391047,
      "eval_runtime": 36.8301,
      "eval_samples_per_second": 760.058,
      "eval_steps_per_second": 11.892,
      "flow/cos_sim": 0.9557033510512958,
      "flow/improvement_ratio": 0.9939037547808259,
      "flow/mag_ratio_mean": 0.92809867015168,
      "flow/mag_ratio_std": 0.10807815949394278,
      "step": 3072
    },
    {
      "epoch": 0.15371114498175603,
      "grad_norm": 0.165770024061203,
      "learning_rate": 0.0009568794565203123,
      "loss": 0.12700329720973969,
      "step": 3328
    },
    {
      "epoch": 0.1655350792111219,
      "grad_norm": 0.17231467366218567,
      "learning_rate": 0.0009488225396630347,
      "loss": 0.12523098289966583,
      "step": 3584
    },
    {
      "epoch": 0.17735901344048774,
      "grad_norm": 0.1858324557542801,
      "learning_rate": 0.0009401160421685646,
      "loss": 0.12655843794345856,
      "step": 3840
    },
    {
      "epoch": 0.18918294766985358,
      "grad_norm": 0.11940455436706543,
      "learning_rate": 0.0009307725649463714,
      "loss": 0.12418897449970245,
      "step": 4096
    },
    {
      "epoch": 0.18918294766985358,
      "eval_cos_loss": 0.04446321669710825,
      "eval_loss": 0.08204046752490954,
      "eval_mse_loss": 0.08204046752490954,
      "flow/cos_sim": 0.9555367968942476,
      "flow/improvement_ratio": 0.9933926257640804,
      "flow/mag_ratio_mean": 0.9328070380916335,
      "flow/mag_ratio_std": 0.10609594972370422,
      "step": 4096
    },
    {
      "epoch": 0.18918294766985358,
      "eval_cos_loss": 0.04446321669710825,
      "eval_loss": 0.08204046752490954,
      "eval_mse_loss": 0.08204046752490954,
      "eval_runtime": 36.8821,
      "eval_samples_per_second": 758.987,
      "eval_steps_per_second": 11.876,
      "flow/cos_sim": 0.9555367968942476,
      "flow/improvement_ratio": 0.9933926257640804,
      "flow/mag_ratio_mean": 0.9328070380916335,
      "flow/mag_ratio_std": 0.10609594972370422,
      "step": 4096
    },
    {
      "epoch": 0.20100688189921945,
      "grad_norm": 0.136522576212883,
      "learning_rate": 0.0009208056308063659,
      "loss": 0.12340260297060013,
      "step": 4352
    },
    {
      "epoch": 0.2128308161285853,
      "grad_norm": 0.18442140519618988,
      "learning_rate": 0.0009102296648873445,
      "loss": 0.12370757013559341,
      "step": 4608
    },
    {
      "epoch": 0.22465475035795113,
      "grad_norm": 0.16970551013946533,
      "learning_rate": 0.0008990599737794927,
      "loss": 0.12366028130054474,
      "step": 4864
    },
    {
      "epoch": 0.236478684587317,
      "grad_norm": 0.13654212653636932,
      "learning_rate": 0.0008873127233711644,
      "loss": 0.12185485661029816,
      "step": 5120
    },
    {
      "epoch": 0.236478684587317,
      "eval_cos_loss": 0.043456991479549234,
      "eval_loss": 0.0809854866945308,
      "eval_mse_loss": 0.0809854866945308,
      "flow/cos_sim": 0.9565430256329714,
      "flow/improvement_ratio": 0.9940090940151041,
      "flow/mag_ratio_mean": 0.9294795245612593,
      "flow/mag_ratio_std": 0.10000736579367014,
      "step": 5120
    },
    {
      "epoch": 0.236478684587317,
      "eval_cos_loss": 0.043456991479549234,
      "eval_loss": 0.0809854866945308,
      "eval_mse_loss": 0.0809854866945308,
      "eval_runtime": 37.5627,
      "eval_samples_per_second": 745.234,
      "eval_steps_per_second": 11.661,
      "flow/cos_sim": 0.9565430256329714,
      "flow/improvement_ratio": 0.9940090940151041,
      "flow/mag_ratio_mean": 0.9294795245612593,
      "flow/mag_ratio_std": 0.10000736579367014,
      "step": 5120
    },
    {
      "epoch": 0.24830261881668284,
      "grad_norm": 0.1422841101884842,
      "learning_rate": 0.0008750049154520011,
      "loss": 0.12234608083963394,
      "step": 5376
    },
    {
      "epoch": 0.2601265530460487,
      "grad_norm": 0.24812348186969757,
      "learning_rate": 0.0008621543631062487,
      "loss": 0.12298071384429932,
      "step": 5632
    },
    {
      "epoch": 0.27195048727541454,
      "grad_norm": 0.16056349873542786,
      "learning_rate": 0.0008487796649318904,
      "loss": 0.12491145730018616,
      "step": 5888
    },
    {
      "epoch": 0.2837744215047804,
      "grad_norm": 0.1158125102519989,
      "learning_rate": 0.0008349001781229053,
      "loss": 0.12146689742803574,
      "step": 6144
    },
    {
      "epoch": 0.2837744215047804,
      "eval_cos_loss": 0.04404348295743335,
      "eval_loss": 0.08143461094191086,
      "eval_mse_loss": 0.08143461094191086,
      "flow/cos_sim": 0.955956531578003,
      "flow/improvement_ratio": 0.9938269595032958,
      "flow/mag_ratio_mean": 0.9270812613234672,
      "flow/mag_ratio_std": 0.10172502801128447,
      "step": 6144
    },
    {
      "epoch": 0.2837744215047804,
      "eval_cos_loss": 0.04404348295743335,
      "eval_loss": 0.08143461094191086,
      "eval_mse_loss": 0.08143461094191086,
      "eval_runtime": 37.2365,
      "eval_samples_per_second": 751.762,
      "eval_steps_per_second": 11.763,
      "flow/cos_sim": 0.955956531578003,
      "flow/improvement_ratio": 0.9938269595032958,
      "flow/mag_ratio_mean": 0.9270812613234672,
      "flow/mag_ratio_std": 0.10172502801128447,
      "step": 6144
    },
    {
      "epoch": 0.2955983557341462,
      "grad_norm": 0.12894846498966217,
      "learning_rate": 0.0008205359904536107,
      "loss": 0.12104764580726624,
      "step": 6400
    },
    {
      "epoch": 0.30742228996351206,
      "grad_norm": 0.1334332525730133,
      "learning_rate": 0.0008057078912056363,
      "loss": 0.12087016552686691,
      "step": 6656
    },
    {
      "epoch": 0.3192462241928779,
      "grad_norm": 0.134778693318367,
      "learning_rate": 0.0007904373410796086,
      "loss": 0.12076316028833389,
      "step": 6912
    },
    {
      "epoch": 0.3310701584222438,
      "grad_norm": 0.11102022230625153,
      "learning_rate": 0.0007747464411350876,
      "loss": 0.12042346596717834,
      "step": 7168
    },
    {
      "epoch": 0.3310701584222438,
      "eval_cos_loss": 0.04294020341938897,
      "eval_loss": 0.07952374037882509,
      "eval_mse_loss": 0.07952374037882509,
      "flow/cos_sim": 0.9570598111032895,
      "flow/improvement_ratio": 0.9943427937488033,
      "flow/mag_ratio_mean": 0.9337734204449065,
      "flow/mag_ratio_std": 0.09960645821676951,
      "step": 7168
    },
    {
      "epoch": 0.3310701584222438,
      "eval_cos_loss": 0.04294020341938897,
      "eval_loss": 0.07952374037882509,
      "eval_mse_loss": 0.07952374037882509,
      "eval_runtime": 37.5482,
      "eval_samples_per_second": 745.522,
      "eval_steps_per_second": 11.665,
      "flow/cos_sim": 0.9570598111032895,
      "flow/improvement_ratio": 0.9943427937488033,
      "flow/mag_ratio_mean": 0.9337734204449065,
      "flow/mag_ratio_std": 0.09960645821676951,
      "step": 7168
    },
    {
      "epoch": 0.34289409265160964,
      "grad_norm": 0.7088926434516907,
      "learning_rate": 0.000758657900803716,
      "loss": 0.12144383788108826,
      "step": 7424
    },
    {
      "epoch": 0.3547180268809755,
      "grad_norm": 0.12531471252441406,
      "learning_rate": 0.000742195005021869,
      "loss": 0.12000911682844162,
      "step": 7680
    },
    {
      "epoch": 0.3665419611103413,
      "grad_norm": 0.12283795326948166,
      "learning_rate": 0.0007253815805303786,
      "loss": 0.12044861167669296,
      "step": 7936
    },
    {
      "epoch": 0.37836589533970716,
      "grad_norm": 0.10739121586084366,
      "learning_rate": 0.0007082419613901028,
      "loss": 0.12056493014097214,
      "step": 8192
    },
    {
      "epoch": 0.37836589533970716,
      "eval_cos_loss": 0.04426976554435954,
      "eval_loss": 0.08079966172984202,
      "eval_mse_loss": 0.08079966172984202,
      "flow/cos_sim": 0.9557302452657865,
      "flow/improvement_ratio": 0.9934861994769475,
      "flow/mag_ratio_mean": 0.9359486946776577,
      "flow/mag_ratio_std": 0.10338757488213174,
      "step": 8192
    },
    {
      "epoch": 0.37836589533970716,
      "eval_cos_loss": 0.04426976554435954,
      "eval_loss": 0.08079966172984202,
      "eval_mse_loss": 0.08079966172984202,
      "eval_runtime": 37.3561,
      "eval_samples_per_second": 749.356,
      "eval_steps_per_second": 11.725,
      "flow/cos_sim": 0.9557302452657865,
      "flow/improvement_ratio": 0.9934861994769475,
      "flow/mag_ratio_mean": 0.9359486946776577,
      "flow/mag_ratio_std": 0.10338757488213174,
      "step": 8192
    },
    {
      "epoch": 0.390189829569073,
      "grad_norm": 0.12729094922542572,
      "learning_rate": 0.0006908009537632514,
      "loss": 0.11962540447711945,
      "step": 8448
    },
    {
      "epoch": 0.4020137637984389,
      "grad_norm": 0.16566377878189087,
      "learning_rate": 0.0006730838000114403,
      "loss": 0.12100031971931458,
      "step": 8704
    },
    {
      "epoch": 0.41383769802780473,
      "grad_norm": 0.14468614757061005,
      "learning_rate": 0.0006551161421624341,
      "loss": 0.12011279165744781,
      "step": 8960
    },
    {
      "epoch": 0.4256616322571706,
      "grad_norm": 0.1833762377500534,
      "learning_rate": 0.0006369239847984517,
      "loss": 0.11977369338274002,
      "step": 9216
    },
    {
      "epoch": 0.4256616322571706,
      "eval_cos_loss": 0.04329353226901598,
      "eval_loss": 0.07977866525145154,
      "eval_mse_loss": 0.07977866525145154,
      "flow/cos_sim": 0.9567064815732442,
      "flow/improvement_ratio": 0.9943556535189555,
      "flow/mag_ratio_mean": 0.9365287250307597,
      "flow/mag_ratio_std": 0.10248907555115822,
      "step": 9216
    },
    {
      "epoch": 0.4256616322571706,
      "eval_cos_loss": 0.04329353226901598,
      "eval_loss": 0.07977866525145154,
      "eval_mse_loss": 0.07977866525145154,
      "eval_runtime": 37.6413,
      "eval_samples_per_second": 743.678,
      "eval_steps_per_second": 11.636,
      "flow/cos_sim": 0.9567064815732442,
      "flow/improvement_ratio": 0.9943556535189555,
      "flow/mag_ratio_mean": 0.9365287250307597,
      "flow/mag_ratio_std": 0.10248907555115822,
      "step": 9216
    }
  ],
  "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
}