{
  "best_metric": 0.31844088435173035,
  "best_model_checkpoint": "uk-mt5-small-gec/checkpoint-1000",
  "epoch": 3.802281368821293,
  "global_step": 1000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.76,
      "learning_rate": 0.0009239543726235742,
      "loss": 1.6591,
      "step": 200
    },
    {
      "epoch": 0.76,
      "eval_google_bleu": 0.41810232396378044,
      "eval_loss": 0.3892889618873596,
      "eval_runtime": 37.9895,
      "eval_samples_per_second": 49.224,
      "eval_steps_per_second": 1.553,
      "step": 200
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.0008479087452471484,
      "loss": 0.4151,
      "step": 400
    },
    {
      "epoch": 1.52,
      "eval_google_bleu": 0.42063669574293594,
      "eval_loss": 0.3223673403263092,
      "eval_runtime": 37.3371,
      "eval_samples_per_second": 50.084,
      "eval_steps_per_second": 1.58,
      "step": 400
    },
    {
      "epoch": 2.28,
      "learning_rate": 0.0007718631178707225,
      "loss": 0.3459,
      "step": 600
    },
    {
      "epoch": 2.28,
      "eval_google_bleu": 0.421165716556752,
      "eval_loss": 0.32651257514953613,
      "eval_runtime": 37.2905,
      "eval_samples_per_second": 50.147,
      "eval_steps_per_second": 1.582,
      "step": 600
    },
    {
      "epoch": 3.04,
      "learning_rate": 0.0006958174904942966,
      "loss": 0.3001,
      "step": 800
    },
    {
      "epoch": 3.04,
      "eval_google_bleu": 0.4227517103515067,
      "eval_loss": 0.3188486397266388,
      "eval_runtime": 37.7047,
      "eval_samples_per_second": 49.596,
      "eval_steps_per_second": 1.565,
      "step": 800
    },
    {
      "epoch": 3.8,
      "learning_rate": 0.0006197718631178707,
      "loss": 0.2519,
      "step": 1000
    },
    {
      "epoch": 3.8,
      "eval_google_bleu": 0.42282362881781355,
      "eval_loss": 0.31844088435173035,
      "eval_runtime": 37.2519,
      "eval_samples_per_second": 50.199,
      "eval_steps_per_second": 1.584,
      "step": 1000
    }
  ],
  "max_steps": 2630,
  "num_train_epochs": 10,
  "total_flos": 2073623000297472.0,
  "trial_name": null,
  "trial_params": null
}