trainer_state.json

{
  "best_global_step": 6,
  "best_metric": 0.6884352564811707,
  "best_model_checkpoint": "tinybert_base_train_kd_wnli/checkpoint-6",
  "epoch": 7.0,
  "eval_steps": 500,
  "global_step": 21,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 3.9158644676208496,
      "learning_rate": 4.933333333333334e-05,
      "loss": 0.8341,
      "step": 3
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.43661971830985913,
      "eval_loss": 0.7070860266685486,
      "eval_runtime": 0.035,
      "eval_samples_per_second": 2029.507,
      "eval_steps_per_second": 28.585,
      "step": 3
    },
    {
      "epoch": 2.0,
      "grad_norm": 1.0790330171585083,
      "learning_rate": 4.8333333333333334e-05,
      "loss": 0.6963,
      "step": 6
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.5633802816901409,
      "eval_loss": 0.6884352564811707,
      "eval_runtime": 0.0351,
      "eval_samples_per_second": 2021.296,
      "eval_steps_per_second": 28.469,
      "step": 6
    },
    {
      "epoch": 3.0,
      "grad_norm": 0.7972629070281982,
      "learning_rate": 4.7333333333333336e-05,
      "loss": 0.6962,
      "step": 9
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.5070422535211268,
      "eval_loss": 0.6948173642158508,
      "eval_runtime": 0.0342,
      "eval_samples_per_second": 2074.016,
      "eval_steps_per_second": 29.211,
      "step": 9
    },
    {
      "epoch": 4.0,
      "grad_norm": 1.001131534576416,
      "learning_rate": 4.633333333333333e-05,
      "loss": 0.6958,
      "step": 12
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.43661971830985913,
      "eval_loss": 0.7016945481300354,
      "eval_runtime": 0.0353,
      "eval_samples_per_second": 2011.303,
      "eval_steps_per_second": 28.328,
      "step": 12
    },
    {
      "epoch": 5.0,
      "grad_norm": 1.215222954750061,
      "learning_rate": 4.5333333333333335e-05,
      "loss": 0.6979,
      "step": 15
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.4647887323943662,
      "eval_loss": 0.7054358720779419,
      "eval_runtime": 0.0347,
      "eval_samples_per_second": 2043.377,
      "eval_steps_per_second": 28.78,
      "step": 15
    },
    {
      "epoch": 6.0,
      "grad_norm": 1.3821132183074951,
      "learning_rate": 4.433333333333334e-05,
      "loss": 0.6946,
      "step": 18
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.4647887323943662,
      "eval_loss": 0.7001540064811707,
      "eval_runtime": 0.0347,
      "eval_samples_per_second": 2047.521,
      "eval_steps_per_second": 28.838,
      "step": 18
    },
    {
      "epoch": 7.0,
      "grad_norm": 0.8882964253425598,
      "learning_rate": 4.3333333333333334e-05,
      "loss": 0.696,
      "step": 21
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.4647887323943662,
      "eval_loss": 0.7013094425201416,
      "eval_runtime": 0.0339,
      "eval_samples_per_second": 2096.724,
      "eval_steps_per_second": 29.531,
      "step": 21
    },
    {
      "epoch": 7.0,
      "step": 21,
      "total_flos": 294408793512960.0,
      "train_loss": 0.7158349355061849,
      "train_runtime": 13.9464,
      "train_samples_per_second": 2276.569,
      "train_steps_per_second": 10.755
    }
  ],
  "logging_steps": 1,
  "max_steps": 150,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 50,
  "save_steps": 500,
  "stateful_callbacks": {
    "EarlyStoppingCallback": {
      "args": {
        "early_stopping_patience": 5,
        "early_stopping_threshold": 0.0
      },
      "attributes": {
        "early_stopping_patience_counter": 5
      }
    },
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 294408793512960.0,
  "train_batch_size": 256,
  "trial_name": null,
  "trial_params": null
}