trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.0,
  "eval_steps": 500,
  "global_step": 56,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.71,
      "grad_norm": 0.4605858623981476,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.41,
      "step": 10
    },
    {
      "epoch": 1.43,
      "grad_norm": 0.6663874387741089,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.4122,
      "step": 20
    },
    {
      "epoch": 2.14,
      "grad_norm": 0.4776056408882141,
      "learning_rate": 3e-06,
      "loss": 0.3811,
      "step": 30
    },
    {
      "epoch": 2.86,
      "grad_norm": 0.6057049632072449,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.4154,
      "step": 40
    },
    {
      "epoch": 3.57,
      "grad_norm": 0.5491490960121155,
      "learning_rate": 5e-06,
      "loss": 0.3939,
      "step": 50
    }
  ],
  "logging_steps": 10,
  "max_steps": 56,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 500,
  "total_flos": 267941471846400.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}