trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.9987442444537464,
  "eval_steps": 500,
  "global_step": 1791,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.8,
      "grad_norm": 0.3945861756801605,
      "learning_rate": 2.4487895716945996e-05,
      "loss": 2.352,
      "step": 477
    },
    {
      "epoch": 1.6,
      "grad_norm": 0.5365868806838989,
      "learning_rate": 1.5605214152700185e-05,
      "loss": 1.8951,
      "step": 954
    },
    {
      "epoch": 2.4,
      "grad_norm": 0.545904278755188,
      "learning_rate": 6.722532588454377e-06,
      "loss": 1.8523,
      "step": 1431
    }
  ],
  "logging_steps": 477,
  "max_steps": 1791,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 5.884617728899154e+17,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}