trainer_state.json

{
  "best_metric": 0.6473702788352966,
  "best_model_checkpoint": "/scratch1/taw2/semeval/nli_upload/checkpoint-162",
  "epoch": 5.0,
  "global_step": 270,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "eval_accuracy": 0.5,
      "eval_f1": 0.0,
      "eval_loss": 0.6974926590919495,
      "eval_precision": 0.0,
      "eval_recall": 0.0,
      "eval_runtime": 0.739,
      "eval_samples_per_second": 270.621,
      "eval_steps_per_second": 9.472,
      "step": 54
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.59,
      "eval_f1": 0.5060240963855421,
      "eval_loss": 0.6564074754714966,
      "eval_precision": 0.6363636363636364,
      "eval_recall": 0.42,
      "eval_runtime": 0.6921,
      "eval_samples_per_second": 288.987,
      "eval_steps_per_second": 10.115,
      "step": 108
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.615,
      "eval_f1": 0.641860465116279,
      "eval_loss": 0.6473702788352966,
      "eval_precision": 0.6,
      "eval_recall": 0.69,
      "eval_runtime": 0.6899,
      "eval_samples_per_second": 289.893,
      "eval_steps_per_second": 10.146,
      "step": 162
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.6,
      "eval_f1": 0.6638655462184875,
      "eval_loss": 0.6594653129577637,
      "eval_precision": 0.572463768115942,
      "eval_recall": 0.79,
      "eval_runtime": 0.6906,
      "eval_samples_per_second": 289.595,
      "eval_steps_per_second": 10.136,
      "step": 216
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.61,
      "eval_f1": 0.6355140186915889,
      "eval_loss": 0.7758804559707642,
      "eval_precision": 0.5964912280701754,
      "eval_recall": 0.68,
      "eval_runtime": 0.6911,
      "eval_samples_per_second": 289.401,
      "eval_steps_per_second": 10.129,
      "step": 270
    },
    {
      "epoch": 5.0,
      "step": 270,
      "total_flos": 837689804211456.0,
      "train_loss": 0.623549058702257,
      "train_runtime": 216.6633,
      "train_samples_per_second": 39.231,
      "train_steps_per_second": 1.246
    }
  ],
  "max_steps": 270,
  "num_train_epochs": 5,
  "total_flos": 837689804211456.0,
  "trial_name": null,
  "trial_params": null
}