{
  "best_metric": 0.9958183765411377,
  "best_model_checkpoint": "miner_id_24/checkpoint-50",
  "epoch": 0.044740227056652315,
  "eval_steps": 25,
  "global_step": 50,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0008948045411330463,
      "grad_norm": 3.836946964263916,
      "learning_rate": 5e-05,
      "loss": 1.5658,
      "step": 1
    },
    {
      "epoch": 0.0008948045411330463,
      "eval_loss": 1.907202959060669,
      "eval_runtime": 3.2967,
      "eval_samples_per_second": 15.167,
      "eval_steps_per_second": 3.943,
      "step": 1
    },
    {
      "epoch": 0.0017896090822660925,
      "grad_norm": 4.166421413421631,
      "learning_rate": 0.0001,
      "loss": 1.6281,
      "step": 2
    },
    {
      "epoch": 0.002684413623399139,
      "grad_norm": 2.6555700302124023,
      "learning_rate": 9.990365154573717e-05,
      "loss": 1.5267,
      "step": 3
    },
    {
      "epoch": 0.003579218164532185,
      "grad_norm": 1.0376940965652466,
      "learning_rate": 9.961501876182148e-05,
      "loss": 1.2068,
      "step": 4
    },
    {
      "epoch": 0.004474022705665231,
      "grad_norm": 1.283631443977356,
      "learning_rate": 9.913533761814537e-05,
      "loss": 1.1342,
      "step": 5
    },
    {
      "epoch": 0.005368827246798278,
      "grad_norm": 0.9173265099525452,
      "learning_rate": 9.846666218300807e-05,
      "loss": 1.0993,
      "step": 6
    },
    {
      "epoch": 0.006263631787931324,
      "grad_norm": 0.6855677366256714,
      "learning_rate": 9.761185582727977e-05,
      "loss": 1.0748,
      "step": 7
    },
    {
      "epoch": 0.00715843632906437,
      "grad_norm": 0.6936066150665283,
      "learning_rate": 9.657457896300791e-05,
      "loss": 1.0153,
      "step": 8
    },
    {
      "epoch": 0.008053240870197416,
      "grad_norm": 0.721575140953064,
      "learning_rate": 9.535927336897098e-05,
      "loss": 0.9656,
      "step": 9
    },
    {
      "epoch": 0.008948045411330463,
      "grad_norm": 0.7673898935317993,
      "learning_rate": 9.397114317029975e-05,
      "loss": 0.9978,
      "step": 10
    },
    {
      "epoch": 0.009842849952463508,
      "grad_norm": 0.7630154490470886,
      "learning_rate": 9.241613255361455e-05,
      "loss": 1.0211,
      "step": 11
    },
    {
      "epoch": 0.010737654493596555,
      "grad_norm": 0.8820596933364868,
      "learning_rate": 9.070090031310558e-05,
      "loss": 1.0437,
      "step": 12
    },
    {
      "epoch": 0.0116324590347296,
      "grad_norm": 0.6509247422218323,
      "learning_rate": 8.883279133655399e-05,
      "loss": 0.998,
      "step": 13
    },
    {
      "epoch": 0.012527263575862648,
      "grad_norm": 0.5232754349708557,
      "learning_rate": 8.681980515339464e-05,
      "loss": 0.9765,
      "step": 14
    },
    {
      "epoch": 0.013422068116995693,
      "grad_norm": 0.5166589617729187,
      "learning_rate": 8.467056167950311e-05,
      "loss": 1.0507,
      "step": 15
    },
    {
      "epoch": 0.01431687265812874,
      "grad_norm": 0.5327579975128174,
      "learning_rate": 8.239426430539243e-05,
      "loss": 0.9647,
      "step": 16
    },
    {
      "epoch": 0.015211677199261786,
      "grad_norm": 0.5838713049888611,
      "learning_rate": 8.000066048588211e-05,
      "loss": 0.9939,
      "step": 17
    },
    {
      "epoch": 0.01610648174039483,
      "grad_norm": 0.5781336426734924,
      "learning_rate": 7.75e-05,
      "loss": 0.9748,
      "step": 18
    },
    {
      "epoch": 0.01700128628152788,
      "grad_norm": 0.6353739500045776,
      "learning_rate": 7.490299105985507e-05,
      "loss": 0.9621,
      "step": 19
    },
    {
      "epoch": 0.017896090822660925,
      "grad_norm": 0.6439753174781799,
      "learning_rate": 7.222075445642904e-05,
      "loss": 1.0062,
      "step": 20
    },
    {
      "epoch": 0.018790895363793973,
      "grad_norm": 0.6834773421287537,
      "learning_rate": 6.946477593864228e-05,
      "loss": 0.9197,
      "step": 21
    },
    {
      "epoch": 0.019685699904927016,
      "grad_norm": 0.6420912146568298,
      "learning_rate": 6.664685702961344e-05,
      "loss": 0.9302,
      "step": 22
    },
    {
      "epoch": 0.020580504446060063,
      "grad_norm": 0.6515995860099792,
      "learning_rate": 6.377906449072578e-05,
      "loss": 0.9076,
      "step": 23
    },
    {
      "epoch": 0.02147530898719311,
      "grad_norm": 0.7485411167144775,
      "learning_rate": 6.087367864990233e-05,
      "loss": 0.9473,
      "step": 24
    },
    {
      "epoch": 0.022370113528326158,
      "grad_norm": 0.773590624332428,
      "learning_rate": 5.794314081535644e-05,
      "loss": 0.8541,
      "step": 25
    },
    {
      "epoch": 0.022370113528326158,
      "eval_loss": 1.0094516277313232,
      "eval_runtime": 3.3963,
      "eval_samples_per_second": 14.722,
      "eval_steps_per_second": 3.828,
      "step": 25
    },
    {
      "epoch": 0.0232649180694592,
      "grad_norm": 0.5019625425338745,
      "learning_rate": 5.500000000000001e-05,
      "loss": 0.9572,
      "step": 26
    },
    {
      "epoch": 0.02415972261059225,
      "grad_norm": 0.5243887901306152,
      "learning_rate": 5.205685918464356e-05,
      "loss": 0.982,
      "step": 27
    },
    {
      "epoch": 0.025054527151725296,
      "grad_norm": 0.5398479104042053,
      "learning_rate": 4.912632135009769e-05,
      "loss": 0.937,
      "step": 28
    },
    {
      "epoch": 0.025949331692858343,
      "grad_norm": 0.5993919968605042,
      "learning_rate": 4.6220935509274235e-05,
      "loss": 0.9682,
      "step": 29
    },
    {
      "epoch": 0.026844136233991386,
      "grad_norm": 0.534296452999115,
      "learning_rate": 4.3353142970386564e-05,
      "loss": 0.9194,
      "step": 30
    },
    {
      "epoch": 0.027738940775124434,
      "grad_norm": 0.5929005742073059,
      "learning_rate": 4.053522406135775e-05,
      "loss": 0.9588,
      "step": 31
    },
    {
      "epoch": 0.02863374531625748,
      "grad_norm": 0.5979087352752686,
      "learning_rate": 3.777924554357096e-05,
      "loss": 0.9557,
      "step": 32
    },
    {
      "epoch": 0.029528549857390528,
      "grad_norm": 0.5713499784469604,
      "learning_rate": 3.509700894014496e-05,
      "loss": 0.9217,
      "step": 33
    },
    {
      "epoch": 0.03042335439852357,
      "grad_norm": 0.6617943048477173,
      "learning_rate": 3.250000000000001e-05,
      "loss": 0.9979,
      "step": 34
    },
    {
      "epoch": 0.03131815893965662,
      "grad_norm": 0.6407080292701721,
      "learning_rate": 2.9999339514117912e-05,
      "loss": 0.9964,
      "step": 35
    },
    {
      "epoch": 0.03221296348078966,
      "grad_norm": 0.6977891325950623,
      "learning_rate": 2.760573569460757e-05,
      "loss": 0.944,
      "step": 36
    },
    {
      "epoch": 0.03310776802192271,
      "grad_norm": 0.6950139403343201,
      "learning_rate": 2.53294383204969e-05,
      "loss": 0.9345,
      "step": 37
    },
    {
      "epoch": 0.03400257256305576,
      "grad_norm": 0.5052092671394348,
      "learning_rate": 2.3180194846605367e-05,
      "loss": 0.9109,
      "step": 38
    },
    {
      "epoch": 0.034897377104188804,
      "grad_norm": 0.47866344451904297,
      "learning_rate": 2.1167208663446025e-05,
      "loss": 0.9576,
      "step": 39
    },
    {
      "epoch": 0.03579218164532185,
      "grad_norm": 0.4459432065486908,
      "learning_rate": 1.9299099686894423e-05,
      "loss": 0.9317,
      "step": 40
    },
    {
      "epoch": 0.0366869861864549,
      "grad_norm": 0.505197286605835,
      "learning_rate": 1.758386744638546e-05,
      "loss": 0.9577,
      "step": 41
    },
    {
      "epoch": 0.037581790727587945,
      "grad_norm": 0.5025796294212341,
      "learning_rate": 1.602885682970026e-05,
      "loss": 0.9043,
      "step": 42
    },
    {
      "epoch": 0.03847659526872099,
      "grad_norm": 0.5344678163528442,
      "learning_rate": 1.464072663102903e-05,
      "loss": 0.9197,
      "step": 43
    },
    {
      "epoch": 0.03937139980985403,
      "grad_norm": 0.5904696583747864,
      "learning_rate": 1.3425421036992098e-05,
      "loss": 0.9859,
      "step": 44
    },
    {
      "epoch": 0.04026620435098708,
      "grad_norm": 0.6027600169181824,
      "learning_rate": 1.2388144172720251e-05,
      "loss": 0.9812,
      "step": 45
    },
    {
      "epoch": 0.04116100889212013,
      "grad_norm": 0.6179023385047913,
      "learning_rate": 1.1533337816991932e-05,
      "loss": 0.9288,
      "step": 46
    },
    {
      "epoch": 0.042055813433253174,
      "grad_norm": 0.6149353981018066,
      "learning_rate": 1.0864662381854632e-05,
      "loss": 0.9331,
      "step": 47
    },
    {
      "epoch": 0.04295061797438622,
      "grad_norm": 0.6846910715103149,
      "learning_rate": 1.0384981238178534e-05,
      "loss": 0.9474,
      "step": 48
    },
    {
      "epoch": 0.04384542251551927,
      "grad_norm": 0.6909250020980835,
      "learning_rate": 1.0096348454262845e-05,
      "loss": 0.925,
      "step": 49
    },
    {
      "epoch": 0.044740227056652315,
      "grad_norm": 0.7580010890960693,
      "learning_rate": 1e-05,
      "loss": 0.938,
      "step": 50
    },
    {
      "epoch": 0.044740227056652315,
      "eval_loss": 0.9958183765411377,
      "eval_runtime": 3.3791,
      "eval_samples_per_second": 14.797,
      "eval_steps_per_second": 3.847,
      "step": 50
    }
  ],
  "logging_steps": 1,
  "max_steps": 50,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 25,
  "stateful_callbacks": {
    "EarlyStoppingCallback": {
      "args": {
        "early_stopping_patience": 1,
        "early_stopping_threshold": 0.0
      },
      "attributes": {
        "early_stopping_patience_counter": 0
      }
    },
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 6.631381723512832e+17,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}