trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 4134,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.36,
      "grad_norm": 0.018075617030262947,
      "learning_rate": 4.3952588292210934e-05,
      "loss": 0.1598,
      "step": 500
    },
    {
      "epoch": 0.73,
      "grad_norm": 0.014405773021280766,
      "learning_rate": 3.790517658442187e-05,
      "loss": 0.0001,
      "step": 1000
    },
    {
      "epoch": 1.09,
      "grad_norm": 0.0018536851275712252,
      "learning_rate": 3.18577648766328e-05,
      "loss": 0.0001,
      "step": 1500
    },
    {
      "epoch": 1.45,
      "grad_norm": 0.0023344324436038733,
      "learning_rate": 2.5810353168843737e-05,
      "loss": 0.0,
      "step": 2000
    },
    {
      "epoch": 1.81,
      "grad_norm": 0.0010399112943559885,
      "learning_rate": 1.9762941461054672e-05,
      "loss": 0.0,
      "step": 2500
    },
    {
      "epoch": 2.18,
      "grad_norm": 0.001136175007559359,
      "learning_rate": 1.3715529753265602e-05,
      "loss": 0.0013,
      "step": 3000
    },
    {
      "epoch": 2.54,
      "grad_norm": 0.001928066136315465,
      "learning_rate": 7.668118045476536e-06,
      "loss": 0.0,
      "step": 3500
    },
    {
      "epoch": 2.9,
      "grad_norm": 0.0015899927821010351,
      "learning_rate": 1.62070633768747e-06,
      "loss": 0.0,
      "step": 4000
    },
    {
      "epoch": 3.0,
      "step": 4134,
      "total_flos": 20904168259584.0,
      "train_loss": 0.01952261976593809,
      "train_runtime": 337.4673,
      "train_samples_per_second": 48.983,
      "train_steps_per_second": 12.25
    }
  ],
  "logging_steps": 500,
  "max_steps": 4134,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 20904168259584.0,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}