trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.021627794311024986,
  "eval_steps": 100000000,
  "global_step": 3000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 1.0000000000000001e-07,
      "loss": 10.7586,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 5e-05,
      "loss": 6.5292,
      "step": 500
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.99998214600225e-05,
      "loss": 4.7262,
      "step": 1000
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.9999285842640105e-05,
      "loss": 4.1997,
      "step": 1500
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.999839315550315e-05,
      "loss": 3.8892,
      "step": 2000
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.9997143411362066e-05,
      "loss": 3.6618,
      "step": 2500
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.9995536628067196e-05,
      "loss": 3.5132,
      "step": 3000
    }
  ],
  "logging_steps": 500,
  "max_steps": 416130,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 3000,
  "total_flos": 1.098714710016e+18,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}