run-14/checkpoint-288/trainer_state.json

{
  "best_metric": 0.8072407045009785,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-14/checkpoint-288",
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 288,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 1.392162561416626,
      "learning_rate": 0.00018327679510936405,
      "loss": 0.506,
      "step": 96
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7945205479452054,
      "eval_f1": 0.8138297872340425,
      "eval_loss": 0.43322718143463135,
      "eval_precision": 0.7439222042139384,
      "eval_recall": 0.898238747553816,
      "eval_runtime": 28.335,
      "eval_samples_per_second": 36.068,
      "eval_steps_per_second": 1.129,
      "step": 96
    },
    {
      "epoch": 2.0,
      "grad_norm": 2.713252544403076,
      "learning_rate": 0.00012218453007290937,
      "loss": 0.4356,
      "step": 192
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.7974559686888454,
      "eval_f1": 0.8169761273209548,
      "eval_loss": 0.4169149696826935,
      "eval_precision": 0.7451612903225806,
      "eval_recall": 0.9041095890410958,
      "eval_runtime": 28.1818,
      "eval_samples_per_second": 36.265,
      "eval_steps_per_second": 1.135,
      "step": 192
    },
    {
      "epoch": 3.0,
      "grad_norm": 2.8020107746124268,
      "learning_rate": 6.109226503645469e-05,
      "loss": 0.4193,
      "step": 288
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8072407045009785,
      "eval_f1": 0.830318690783807,
      "eval_loss": 0.4170421361923218,
      "eval_precision": 0.7415384615384616,
      "eval_recall": 0.9432485322896281,
      "eval_runtime": 28.2719,
      "eval_samples_per_second": 36.149,
      "eval_steps_per_second": 1.132,
      "step": 288
    }
  ],
  "logging_steps": 500,
  "max_steps": 384,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 500,
  "total_flos": 707085591840.0,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.7512024699670536,
    "learning_rate": 0.00024436906014581875,
    "num_train_epochs": 4,
    "temperature": 25
  }
}