last-checkpoint/trainer_state.json

{
  "best_metric": 125.62500000000001,
  "best_model_checkpoint": "../openai/whisper-small-finetuned/checkpoint-25",
  "epoch": 33.01,
  "eval_steps": 25,
  "global_step": 100,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 8.01,
      "grad_norm": 12.496959686279297,
      "learning_rate": 7.7e-06,
      "loss": 4.7293,
      "step": 25
    },
    {
      "epoch": 8.01,
      "eval_loss": 5.816229820251465,
      "eval_runtime": 28.1944,
      "eval_samples_per_second": 0.709,
      "eval_steps_per_second": 0.035,
      "eval_wer": 125.62500000000001,
      "step": 25
    },
    {
      "epoch": 16.02,
      "grad_norm": 6.95028018951416,
      "learning_rate": 5.2e-06,
      "loss": 4.1734,
      "step": 50
    },
    {
      "epoch": 16.02,
      "eval_loss": 5.398538589477539,
      "eval_runtime": 27.3042,
      "eval_samples_per_second": 0.732,
      "eval_steps_per_second": 0.037,
      "eval_wer": 130.0,
      "step": 50
    },
    {
      "epoch": 24.03,
      "grad_norm": 4.860121250152588,
      "learning_rate": 2.7000000000000004e-06,
      "loss": 3.8912,
      "step": 75
    },
    {
      "epoch": 24.03,
      "eval_loss": 5.164035797119141,
      "eval_runtime": 27.1811,
      "eval_samples_per_second": 0.736,
      "eval_steps_per_second": 0.037,
      "eval_wer": 125.62500000000001,
      "step": 75
    },
    {
      "epoch": 33.01,
      "grad_norm": 2.9855027198791504,
      "learning_rate": 2.0000000000000002e-07,
      "loss": 3.7045,
      "step": 100
    },
    {
      "epoch": 33.01,
      "eval_loss": 5.087833404541016,
      "eval_runtime": 27.4776,
      "eval_samples_per_second": 0.728,
      "eval_steps_per_second": 0.036,
      "eval_wer": 128.125,
      "step": 100
    }
  ],
  "logging_steps": 25,
  "max_steps": 100,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 9223372036854775807,
  "save_steps": 25,
  "total_flos": 1.96649118793728e+18,
  "train_batch_size": 96,
  "trial_name": null,
  "trial_params": null
}