trainer_state.json

{
  "best_global_step": 820,
  "best_metric": 0.3988942801952362,
  "best_model_checkpoint": "tinybert_base_train_kd_qnli/checkpoint-820",
  "epoch": 7.0,
  "eval_steps": 500,
  "global_step": 2870,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 4.125197410583496,
      "learning_rate": 4.900243902439024e-05,
      "loss": 0.5274,
      "step": 410
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7980962840929892,
      "eval_loss": 0.4534500241279602,
      "eval_runtime": 2.6921,
      "eval_samples_per_second": 2029.264,
      "eval_steps_per_second": 8.172,
      "step": 410
    },
    {
      "epoch": 2.0,
      "grad_norm": 4.953464508056641,
      "learning_rate": 4.8002439024390246e-05,
      "loss": 0.4315,
      "step": 820
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8184147904082006,
      "eval_loss": 0.3988942801952362,
      "eval_runtime": 2.7087,
      "eval_samples_per_second": 2016.83,
      "eval_steps_per_second": 8.122,
      "step": 820
    },
    {
      "epoch": 3.0,
      "grad_norm": 4.556070804595947,
      "learning_rate": 4.700243902439024e-05,
      "loss": 0.3719,
      "step": 1230
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.7966318872414424,
      "eval_loss": 0.4447864890098572,
      "eval_runtime": 2.7383,
      "eval_samples_per_second": 1995.067,
      "eval_steps_per_second": 8.034,
      "step": 1230
    },
    {
      "epoch": 4.0,
      "grad_norm": 5.391785144805908,
      "learning_rate": 4.600243902439025e-05,
      "loss": 0.3144,
      "step": 1640
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.8123741533955702,
      "eval_loss": 0.43414977192878723,
      "eval_runtime": 2.7559,
      "eval_samples_per_second": 1982.316,
      "eval_steps_per_second": 7.983,
      "step": 1640
    },
    {
      "epoch": 5.0,
      "grad_norm": 5.456552505493164,
      "learning_rate": 4.500243902439025e-05,
      "loss": 0.266,
      "step": 2050
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.823906278601501,
      "eval_loss": 0.4281355142593384,
      "eval_runtime": 2.6884,
      "eval_samples_per_second": 2032.035,
      "eval_steps_per_second": 8.183,
      "step": 2050
    },
    {
      "epoch": 6.0,
      "grad_norm": 7.007852554321289,
      "learning_rate": 4.400243902439024e-05,
      "loss": 0.218,
      "step": 2460
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.8228079809628409,
      "eval_loss": 0.5215159058570862,
      "eval_runtime": 2.6375,
      "eval_samples_per_second": 2071.27,
      "eval_steps_per_second": 8.341,
      "step": 2460
    },
    {
      "epoch": 7.0,
      "grad_norm": 3.821591377258301,
      "learning_rate": 4.3002439024390246e-05,
      "loss": 0.1743,
      "step": 2870
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.8207944352919642,
      "eval_loss": 0.5190011262893677,
      "eval_runtime": 2.7166,
      "eval_samples_per_second": 2010.967,
      "eval_steps_per_second": 8.098,
      "step": 2870
    },
    {
      "epoch": 7.0,
      "step": 2870,
      "total_flos": 4.856261458098893e+16,
      "train_loss": 0.32907051591507647,
      "train_runtime": 728.71,
      "train_samples_per_second": 7186.878,
      "train_steps_per_second": 28.132
    }
  ],
  "logging_steps": 1,
  "max_steps": 20500,
  "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": 4.856261458098893e+16,
  "train_batch_size": 256,
  "trial_name": null,
  "trial_params": null
}