run-0/checkpoint-2526/trainer_state.json

{
  "best_metric": 0.8990825688073395,
  "best_model_checkpoint": "bert-base-uncased-finetuned-sst2/run-0/checkpoint-2526",
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 2526,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.59,
      "grad_norm": 14.076563835144043,
      "learning_rate": 2.1488708627178514e-06,
      "loss": 0.4976,
      "step": 500
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8646788990825688,
      "eval_loss": 0.33198288083076477,
      "eval_runtime": 2.248,
      "eval_samples_per_second": 387.892,
      "eval_steps_per_second": 24.466,
      "step": 842
    },
    {
      "epoch": 1.19,
      "grad_norm": 16.497121810913086,
      "learning_rate": 1.8592656251547983e-06,
      "loss": 0.3072,
      "step": 1000
    },
    {
      "epoch": 1.78,
      "grad_norm": 50.02860641479492,
      "learning_rate": 1.5696603875917459e-06,
      "loss": 0.3036,
      "step": 1500
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8876146788990825,
      "eval_loss": 0.3288116455078125,
      "eval_runtime": 2.3363,
      "eval_samples_per_second": 373.233,
      "eval_steps_per_second": 23.541,
      "step": 1684
    },
    {
      "epoch": 2.38,
      "grad_norm": 0.35839420557022095,
      "learning_rate": 1.280055150028693e-06,
      "loss": 0.2703,
      "step": 2000
    },
    {
      "epoch": 2.97,
      "grad_norm": 65.75403594970703,
      "learning_rate": 9.904499124656404e-07,
      "loss": 0.2546,
      "step": 2500
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8990825688073395,
      "eval_loss": 0.3764495253562927,
      "eval_runtime": 2.384,
      "eval_samples_per_second": 365.774,
      "eval_steps_per_second": 23.071,
      "step": 2526
    }
  ],
  "logging_steps": 500,
  "max_steps": 4210,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 306837851650200.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": {
    "learning_rate": 2.438476100280904e-06,
    "num_train_epochs": 5,
    "per_device_train_batch_size": 8,
    "seed": 29
  }
}