run-36/checkpoint-642/trainer_state.json

{
  "best_metric": 0.7305213825424721,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-36/checkpoint-642",
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 642,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 9.520071029663086,
      "learning_rate": 0.00017796815168800775,
      "loss": 0.6005,
      "step": 214
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7070884592852958,
      "eval_f1": 0.4369369369369369,
      "eval_loss": 0.5579986572265625,
      "eval_mcc": 0.2794799623456291,
      "eval_precision": 0.6081504702194357,
      "eval_recall": 0.3409490333919156,
      "eval_runtime": 3.1263,
      "eval_samples_per_second": 546.008,
      "eval_steps_per_second": 17.273,
      "step": 214
    },
    {
      "epoch": 2.0,
      "grad_norm": 2.303219795227051,
      "learning_rate": 0.00015819391261156243,
      "loss": 0.5376,
      "step": 428
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.7287639132981839,
      "eval_f1": 0.5141657922350472,
      "eval_loss": 0.5312134027481079,
      "eval_mcc": 0.3482142857142857,
      "eval_precision": 0.6380208333333334,
      "eval_recall": 0.4305799648506151,
      "eval_runtime": 3.7151,
      "eval_samples_per_second": 459.482,
      "eval_steps_per_second": 14.535,
      "step": 428
    },
    {
      "epoch": 3.0,
      "grad_norm": 2.198146343231201,
      "learning_rate": 0.00013841967353511711,
      "loss": 0.5111,
      "step": 642
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.7305213825424721,
      "eval_f1": 0.42643391521197005,
      "eval_loss": 0.5205321311950684,
      "eval_mcc": 0.33784409399973486,
      "eval_precision": 0.7339055793991416,
      "eval_recall": 0.30052724077328646,
      "eval_runtime": 3.1478,
      "eval_samples_per_second": 542.282,
      "eval_steps_per_second": 17.155,
      "step": 642
    }
  ],
  "logging_steps": 500,
  "max_steps": 2140,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 500,
  "total_flos": 1574326993320.0,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.9927925943509046,
    "learning_rate": 0.00019774239076445304,
    "num_train_epochs": 10,
    "temperature": 21
  }
}