run-27/checkpoint-576/trainer_state.json

{
  "best_metric": 0.8140900195694716,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-27/checkpoint-576",
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 576,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 5.356992244720459,
      "learning_rate": 8.28536608591822e-05,
      "loss": 0.547,
      "step": 192
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.799412915851272,
      "eval_f1": 0.820647419072616,
      "eval_loss": 0.4454892873764038,
      "eval_mcc": 0.6163545180480211,
      "eval_precision": 0.7420886075949367,
      "eval_recall": 0.9178082191780822,
      "eval_runtime": 66.9011,
      "eval_samples_per_second": 15.276,
      "eval_steps_per_second": 0.478,
      "step": 192
    },
    {
      "epoch": 2.0,
      "grad_norm": 5.9460296630859375,
      "learning_rate": 4.14268304295911e-05,
      "loss": 0.4534,
      "step": 384
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8111545988258317,
      "eval_f1": 0.8237442922374429,
      "eval_loss": 0.42414745688438416,
      "eval_mcc": 0.6287581698708938,
      "eval_precision": 0.7722602739726028,
      "eval_recall": 0.8825831702544031,
      "eval_runtime": 66.3484,
      "eval_samples_per_second": 15.404,
      "eval_steps_per_second": 0.482,
      "step": 384
    },
    {
      "epoch": 3.0,
      "grad_norm": 4.355434894561768,
      "learning_rate": 0.0,
      "loss": 0.4285,
      "step": 576
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8140900195694716,
      "eval_f1": 0.8297491039426522,
      "eval_loss": 0.4249483346939087,
      "eval_mcc": 0.6390860201862151,
      "eval_precision": 0.7652892561983471,
      "eval_recall": 0.9060665362035225,
      "eval_runtime": 66.8424,
      "eval_samples_per_second": 15.29,
      "eval_steps_per_second": 0.479,
      "step": 576
    }
  ],
  "logging_steps": 500,
  "max_steps": 576,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 707085591840.0,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.8507816113188046,
    "learning_rate": 0.0001242804912887733,
    "num_train_epochs": 3,
    "per_device_train_batch_size": 16,
    "temperature": 24
  }
}