trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.0,
  "global_step": 3436,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.29,
      "learning_rate": 0.0001286654671352585,
      "loss": 1.9284,
      "step": 500
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00010691674043896736,
      "loss": 1.3223,
      "step": 1000
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.516801374267618e-05,
      "loss": 1.1763,
      "step": 1500
    },
    {
      "epoch": 1.16,
      "learning_rate": 6.341928704638503e-05,
      "loss": 0.9532,
      "step": 2000
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.167056035009386e-05,
      "loss": 0.8007,
      "step": 2500
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.9921833653802702e-05,
      "loss": 0.7703,
      "step": 3000
    },
    {
      "epoch": 2.0,
      "step": 3436,
      "total_flos": 62642546802688.0,
      "train_loss": 1.103950074165886,
      "train_runtime": 518.3768,
      "train_samples_per_second": 503.506,
      "train_steps_per_second": 6.628
    }
  ],
  "max_steps": 3436,
  "num_train_epochs": 2,
  "total_flos": 62642546802688.0,
  "trial_name": null,
  "trial_params": null
}