run-4/checkpoint-144/trainer_state.json

{
  "best_metric": 0.5655577299412916,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-4/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": 0.8744632005691528,
      "learning_rate": 3.33370769821791e-05,
      "loss": 0.5437,
      "step": 48
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.5,
      "eval_f1": 0.0,
      "eval_loss": 0.5369552969932556,
      "eval_precision": 0.0,
      "eval_recall": 0.0,
      "eval_runtime": 28.2527,
      "eval_samples_per_second": 36.174,
      "eval_steps_per_second": 1.133,
      "step": 48
    },
    {
      "epoch": 2.0,
      "grad_norm": 1.112693428993225,
      "learning_rate": 2.2224717988119397e-05,
      "loss": 0.5327,
      "step": 96
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.5195694716242661,
      "eval_f1": 0.11531531531531532,
      "eval_loss": 0.5237143635749817,
      "eval_precision": 0.7272727272727273,
      "eval_recall": 0.06262230919765166,
      "eval_runtime": 28.2522,
      "eval_samples_per_second": 36.174,
      "eval_steps_per_second": 1.133,
      "step": 96
    },
    {
      "epoch": 3.0,
      "grad_norm": 1.7103320360183716,
      "learning_rate": 1.1112358994059699e-05,
      "loss": 0.5201,
      "step": 144
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.5655577299412916,
      "eval_f1": 0.2861736334405145,
      "eval_loss": 0.5045989751815796,
      "eval_precision": 0.8018018018018018,
      "eval_recall": 0.17416829745596868,
      "eval_runtime": 28.4752,
      "eval_samples_per_second": 35.891,
      "eval_steps_per_second": 1.124,
      "step": 144
    }
  ],
  "logging_steps": 500,
  "max_steps": 192,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 500,
  "total_flos": 707085591840.0,
  "train_batch_size": 64,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.7223881376780374,
    "learning_rate": 4.4449435976238795e-05,
    "num_train_epochs": 4,
    "per_device_train_batch_size": 64,
    "temperature": 22
  }
}