trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.88,
  "eval_steps": 500,
  "global_step": 18,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.8,
      "grad_norm": 2.860818386077881,
      "learning_rate": 4.1069690242163484e-05,
      "loss": 7.2633,
      "step": 5
    },
    {
      "epoch": 1.6,
      "grad_norm": 4.06754207611084,
      "learning_rate": 2.0658795558326743e-05,
      "loss": 6.3067,
      "step": 10
    },
    {
      "epoch": 2.4,
      "grad_norm": 4.378876686096191,
      "learning_rate": 3.3493649053890326e-06,
      "loss": 6.4305,
      "step": 15
    },
    {
      "epoch": 2.88,
      "step": 18,
      "total_flos": 133622157312000.0,
      "train_loss": 6.60821893480089,
      "train_runtime": 59.777,
      "train_samples_per_second": 5.019,
      "train_steps_per_second": 0.301
    }
  ],
  "logging_steps": 5,
  "max_steps": 18,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 100,
  "total_flos": 133622157312000.0,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}