checkpoint-2455/trainer_state.json

{
  "best_metric": 0.1743677854537964,
  "best_model_checkpoint": "/home/paperspace/microsoft_deberta-v3-small_cpt_f6d5daab/checkpoint-2455",
  "epoch": 1.0,
  "eval_steps": 500,
  "global_step": 2455,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.2,
      "learning_rate": 9.709048588885654e-07,
      "loss": 0.5815,
      "step": 500
    },
    {
      "epoch": 0.41,
      "learning_rate": 9.418097177771311e-07,
      "loss": 0.2425,
      "step": 1000
    },
    {
      "epoch": 0.61,
      "learning_rate": 9.127145766656967e-07,
      "loss": 0.193,
      "step": 1500
    },
    {
      "epoch": 0.81,
      "learning_rate": 8.836194355542625e-07,
      "loss": 0.175,
      "step": 2000
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.9521287431248727,
      "eval_f1": 0.952353470322894,
      "eval_loss": 0.1743677854537964,
      "eval_precision": 0.9532059236996064,
      "eval_recall": 0.9521287431248727,
      "eval_runtime": 3.5334,
      "eval_samples_per_second": 1389.316,
      "eval_steps_per_second": 86.885,
      "step": 2455
    }
  ],
  "logging_steps": 500,
  "max_steps": 17185,
  "num_train_epochs": 7,
  "save_steps": 500,
  "total_flos": 479115425241456.0,
  "trial_name": null,
  "trial_params": null
}