trainer_state.json

{
  "best_metric": 0.9789915966386554,
  "best_model_checkpoint": "/mnt/data02/luca_newswire/wire_classifier/trained_models/wire/2024-05-29_23-47-53/checkpoint-150",
  "epoch": 6.521739130434782,
  "global_step": 150,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.17,
      "eval_accuracy": 0.7023809523809523,
      "eval_f1": 0.8251748251748251,
      "eval_loss": 0.542344331741333,
      "eval_precision": 0.7023809523809523,
      "eval_recall": 1.0,
      "eval_runtime": 1.9215,
      "eval_samples_per_second": 174.867,
      "eval_steps_per_second": 3.123,
      "step": 50
    },
    {
      "epoch": 4.35,
      "eval_accuracy": 0.8988095238095238,
      "eval_f1": 0.9317269076305221,
      "eval_loss": 0.3049139976501465,
      "eval_precision": 0.8854961832061069,
      "eval_recall": 0.9830508474576272,
      "eval_runtime": 1.6739,
      "eval_samples_per_second": 200.728,
      "eval_steps_per_second": 3.584,
      "step": 100
    },
    {
      "epoch": 6.52,
      "eval_accuracy": 0.9702380952380952,
      "eval_f1": 0.9789915966386554,
      "eval_loss": 0.11797040700912476,
      "eval_precision": 0.9708333333333333,
      "eval_recall": 0.9872881355932204,
      "eval_runtime": 1.6614,
      "eval_samples_per_second": 202.241,
      "eval_steps_per_second": 3.611,
      "step": 150
    }
  ],
  "max_steps": 460,
  "num_train_epochs": 20,
  "total_flos": 1261354570002432.0,
  "trial_name": null,
  "trial_params": null
}