{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 7.372654155495979,
  "eval_steps": 500,
  "global_step": 5500,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.6702412868632708,
      "grad_norm": 6.84100866317749,
      "learning_rate": 2.923321801244596e-06,
      "loss": 0.5942,
      "step": 500
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7338250279426575,
      "eval_loss": 0.5040128231048584,
      "eval_runtime": 8.8539,
      "eval_samples_per_second": 336.913,
      "eval_steps_per_second": 21.121,
      "step": 746
    },
    {
      "epoch": 1.3404825737265416,
      "grad_norm": 10.305294036865234,
      "learning_rate": 2.7133130511551857e-06,
      "loss": 0.5078,
      "step": 1000
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.7418705821037292,
      "eval_loss": 0.4992324709892273,
      "eval_runtime": 8.9157,
      "eval_samples_per_second": 334.577,
      "eval_steps_per_second": 20.974,
      "step": 1492
    },
    {
      "epoch": 2.0107238605898123,
      "grad_norm": 5.501172065734863,
      "learning_rate": 2.5033043010657747e-06,
      "loss": 0.4546,
      "step": 1500
    },
    {
      "epoch": 2.680965147453083,
      "grad_norm": 7.48345947265625,
      "learning_rate": 2.293295550976364e-06,
      "loss": 0.3872,
      "step": 2000
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.7529332637786865,
      "eval_loss": 0.5125850439071655,
      "eval_runtime": 8.8675,
      "eval_samples_per_second": 336.395,
      "eval_steps_per_second": 21.088,
      "step": 2238
    },
    {
      "epoch": 3.351206434316354,
      "grad_norm": 8.878561973571777,
      "learning_rate": 2.0832868008869536e-06,
      "loss": 0.3551,
      "step": 2500
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.750921905040741,
      "eval_loss": 0.5793744921684265,
      "eval_runtime": 8.9125,
      "eval_samples_per_second": 334.698,
      "eval_steps_per_second": 20.982,
      "step": 2984
    },
    {
      "epoch": 4.021447721179625,
      "grad_norm": 17.35097312927246,
      "learning_rate": 1.8732780507975428e-06,
      "loss": 0.3097,
      "step": 3000
    },
    {
      "epoch": 4.6916890080428955,
      "grad_norm": 6.471634387969971,
      "learning_rate": 1.6632693007081323e-06,
      "loss": 0.2532,
      "step": 3500
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.7519276142120361,
      "eval_loss": 0.623068630695343,
      "eval_runtime": 8.8841,
      "eval_samples_per_second": 335.768,
      "eval_steps_per_second": 21.049,
      "step": 3730
    },
    {
      "epoch": 5.361930294906166,
      "grad_norm": 14.375243186950684,
      "learning_rate": 1.4532605506187215e-06,
      "loss": 0.2409,
      "step": 4000
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.7522628307342529,
      "eval_loss": 0.7017402052879333,
      "eval_runtime": 8.8804,
      "eval_samples_per_second": 335.907,
      "eval_steps_per_second": 21.058,
      "step": 4476
    },
    {
      "epoch": 6.032171581769437,
      "grad_norm": 10.026714324951172,
      "learning_rate": 1.243251800529311e-06,
      "loss": 0.2116,
      "step": 4500
    },
    {
      "epoch": 6.702412868632708,
      "grad_norm": 10.323859214782715,
      "learning_rate": 1.0332430504399004e-06,
      "loss": 0.1882,
      "step": 5000
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.7539389729499817,
      "eval_loss": 0.7335813045501709,
      "eval_runtime": 8.8605,
      "eval_samples_per_second": 336.664,
      "eval_steps_per_second": 21.105,
      "step": 5222
    },
    {
      "epoch": 7.372654155495979,
      "grad_norm": 24.86555290222168,
      "learning_rate": 8.232343003504898e-07,
      "loss": 0.1679,
      "step": 5500
    }
  ],
  "logging_steps": 500,
  "max_steps": 7460,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 500,
  "total_flos": 1.307182215348216e+16,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": {
    "learning_rate": 3.133330551334007e-06,
    "per_device_train_batch_size": 16
  }
}