trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.6,
  "eval_steps": 500,
  "global_step": 100,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.4,
      "grad_norm": 16.487810134887695,
      "learning_rate": 4.600000000000001e-05,
      "loss": 2.5731,
      "step": 25
    },
    {
      "epoch": 0.8,
      "grad_norm": 14.608394622802734,
      "learning_rate": 9.6e-05,
      "loss": 0.7107,
      "step": 50
    },
    {
      "epoch": 1.2,
      "grad_norm": 12.708142280578613,
      "learning_rate": 6.891891891891892e-05,
      "loss": 0.5899,
      "step": 75
    },
    {
      "epoch": 1.6,
      "grad_norm": 12.00189208984375,
      "learning_rate": 3.513513513513514e-05,
      "loss": 0.4792,
      "step": 100
    }
  ],
  "logging_steps": 25,
  "max_steps": 124,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 2,
  "save_steps": 25,
  "total_flos": 8017574019985728.0,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}