trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.9492168960607499,
  "eval_steps": 500,
  "global_step": 2000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.24,
      "grad_norm": 2.33807110786438,
      "learning_rate": 3.7500000000000003e-05,
      "loss": 1.1754,
      "step": 500
    },
    {
      "epoch": 0.47,
      "grad_norm": 1.748302698135376,
      "learning_rate": 2.5e-05,
      "loss": 0.9536,
      "step": 1000
    },
    {
      "epoch": 0.71,
      "grad_norm": 1.8499162197113037,
      "learning_rate": 1.25e-05,
      "loss": 0.877,
      "step": 1500
    },
    {
      "epoch": 0.95,
      "grad_norm": 1.7708240747451782,
      "learning_rate": 0.0,
      "loss": 0.8539,
      "step": 2000
    },
    {
      "epoch": 0.95,
      "step": 2000,
      "total_flos": 7429605690441728.0,
      "train_loss": 0.9649612274169922,
      "train_runtime": 2280.9027,
      "train_samples_per_second": 1.754,
      "train_steps_per_second": 0.877
    }
  ],
  "logging_steps": 500,
  "max_steps": 2000,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 1000,
  "total_flos": 7429605690441728.0,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}