trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 10.0,
  "eval_steps": 500,
  "global_step": 3250,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.54,
      "learning_rate": 4.230769230769231e-05,
      "loss": 0.1361,
      "step": 500
    },
    {
      "epoch": 3.08,
      "learning_rate": 3.461538461538462e-05,
      "loss": 0.0416,
      "step": 1000
    },
    {
      "epoch": 4.62,
      "learning_rate": 2.6923076923076923e-05,
      "loss": 0.0154,
      "step": 1500
    },
    {
      "epoch": 6.15,
      "learning_rate": 1.923076923076923e-05,
      "loss": 0.0071,
      "step": 2000
    },
    {
      "epoch": 7.69,
      "learning_rate": 1.153846153846154e-05,
      "loss": 0.0027,
      "step": 2500
    },
    {
      "epoch": 9.23,
      "learning_rate": 3.846153846153847e-06,
      "loss": 0.0021,
      "step": 3000
    },
    {
      "epoch": 10.0,
      "step": 3250,
      "total_flos": 34736444859384.0,
      "train_loss": 0.031623102454038765,
      "train_runtime": 93.4151,
      "train_samples_per_second": 277.792,
      "train_steps_per_second": 34.791
    }
  ],
  "logging_steps": 500,
  "max_steps": 3250,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 500,
  "total_flos": 34736444859384.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}