run-1/checkpoint-144/trainer_state.json

{
  "best_metric": 0.7954990215264188,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-1/checkpoint-144",
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 144,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 1.0160187482833862,
      "learning_rate": 0.00031017361242620324,
      "loss": 0.5046,
      "step": 48
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7338551859099804,
      "eval_f1": 0.7399617590822181,
      "eval_loss": 0.448641836643219,
      "eval_precision": 0.7233644859813084,
      "eval_recall": 0.7573385518590998,
      "eval_runtime": 30.0153,
      "eval_samples_per_second": 34.049,
      "eval_steps_per_second": 0.533,
      "step": 48
    },
    {
      "epoch": 2.0,
      "grad_norm": 3.3045527935028076,
      "learning_rate": 0.00015508680621310162,
      "loss": 0.4485,
      "step": 96
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.7504892367906066,
      "eval_f1": 0.7395301327885597,
      "eval_loss": 0.4257463812828064,
      "eval_precision": 0.7735042735042735,
      "eval_recall": 0.7084148727984344,
      "eval_runtime": 28.6578,
      "eval_samples_per_second": 35.662,
      "eval_steps_per_second": 0.558,
      "step": 96
    },
    {
      "epoch": 3.0,
      "grad_norm": 2.6100542545318604,
      "learning_rate": 0.0,
      "loss": 0.4249,
      "step": 144
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.7954990215264188,
      "eval_f1": 0.8098271155595996,
      "eval_loss": 0.41249141097068787,
      "eval_precision": 0.7568027210884354,
      "eval_recall": 0.8708414872798435,
      "eval_runtime": 29.9599,
      "eval_samples_per_second": 34.112,
      "eval_steps_per_second": 0.534,
      "step": 144
    }
  ],
  "logging_steps": 500,
  "max_steps": 144,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 707085591840.0,
  "train_batch_size": 64,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.7333074246521317,
    "learning_rate": 0.00046526041863930486,
    "num_train_epochs": 3,
    "temperature": 17
  }
}