trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 20.0,
  "eval_steps": 500,
  "global_step": 40,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.5,
      "grad_norm": 37.25093460083008,
      "learning_rate": 4.375e-05,
      "loss": 2.2626,
      "step": 5
    },
    {
      "epoch": 5.0,
      "grad_norm": 49.72227096557617,
      "learning_rate": 3.7500000000000003e-05,
      "loss": 0.8291,
      "step": 10
    },
    {
      "epoch": 7.5,
      "grad_norm": 16.81373405456543,
      "learning_rate": 3.125e-05,
      "loss": 0.8602,
      "step": 15
    },
    {
      "epoch": 10.0,
      "grad_norm": 7.272302627563477,
      "learning_rate": 2.5e-05,
      "loss": 0.5567,
      "step": 20
    },
    {
      "epoch": 12.5,
      "grad_norm": 9.128313064575195,
      "learning_rate": 1.8750000000000002e-05,
      "loss": 0.7576,
      "step": 25
    },
    {
      "epoch": 15.0,
      "grad_norm": 14.296534538269043,
      "learning_rate": 1.25e-05,
      "loss": 0.4636,
      "step": 30
    },
    {
      "epoch": 17.5,
      "grad_norm": 5.532127380371094,
      "learning_rate": 6.25e-06,
      "loss": 0.6616,
      "step": 35
    },
    {
      "epoch": 20.0,
      "grad_norm": 0.8484093546867371,
      "learning_rate": 0.0,
      "loss": 0.3679,
      "step": 40
    },
    {
      "epoch": 20.0,
      "step": 40,
      "total_flos": 87250496949000.0,
      "train_loss": 0.8449213564395904,
      "train_runtime": 26.3276,
      "train_samples_per_second": 56.974,
      "train_steps_per_second": 1.519
    }
  ],
  "logging_steps": 5,
  "max_steps": 40,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 20,
  "save_steps": 500,
  "total_flos": 87250496949000.0,
  "train_batch_size": 64,
  "trial_name": null,
  "trial_params": null
}