trainer_state.json

{
  "best_metric": 0.9393776764051477,
  "best_model_checkpoint": "model_saves/roberta-large_lemon_5k_5_p3/checkpoint-804",
  "epoch": 6.0,
  "global_step": 1608,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "eval_accuracy": 0.9391316986996724,
      "eval_loss": 0.4297237694263458,
      "eval_runtime": 3.3812,
      "eval_samples_per_second": 1296.007,
      "eval_steps_per_second": 10.351,
      "step": 268
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.9390254810541263,
      "eval_loss": 0.44621580839157104,
      "eval_runtime": 3.3932,
      "eval_samples_per_second": 1291.393,
      "eval_steps_per_second": 10.315,
      "step": 536
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.9393776764051477,
      "eval_loss": 0.476402223110199,
      "eval_runtime": 3.3858,
      "eval_samples_per_second": 1294.233,
      "eval_steps_per_second": 10.337,
      "step": 804
    },
    {
      "epoch": 3.73,
      "learning_rate": 1e-05,
      "loss": 0.2902,
      "step": 1000
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.9388186361654312,
      "eval_loss": 0.5053029656410217,
      "eval_runtime": 3.3895,
      "eval_samples_per_second": 1292.82,
      "eval_steps_per_second": 10.326,
      "step": 1072
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.9378403157459274,
      "eval_loss": 0.5688967108726501,
      "eval_runtime": 3.3816,
      "eval_samples_per_second": 1295.838,
      "eval_steps_per_second": 10.35,
      "step": 1340
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.9384888024239985,
      "eval_loss": 0.6369710564613342,
      "eval_runtime": 3.386,
      "eval_samples_per_second": 1294.168,
      "eval_steps_per_second": 10.337,
      "step": 1608
    },
    {
      "epoch": 6.0,
      "step": 1608,
      "total_flos": 2.0794742098886656e+16,
      "train_loss": 0.2427593582305149,
      "train_runtime": 487.3051,
      "train_samples_per_second": 1055.93,
      "train_steps_per_second": 8.249
    }
  ],
  "max_steps": 4020,
  "num_train_epochs": 15,
  "total_flos": 2.0794742098886656e+16,
  "trial_name": null,
  "trial_params": null
}