trainer_state.json

{
  "best_global_step": 528,
  "best_metric": 0.40144169330596924,
  "best_model_checkpoint": "tinybert_base_train_kd_sst2/checkpoint-528",
  "epoch": 7.0,
  "eval_steps": 500,
  "global_step": 1848,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 13.377687454223633,
      "learning_rate": 4.9003787878787885e-05,
      "loss": 0.3628,
      "step": 264
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8291284403669725,
      "eval_loss": 0.41989806294441223,
      "eval_runtime": 0.4445,
      "eval_samples_per_second": 1961.561,
      "eval_steps_per_second": 8.998,
      "step": 264
    },
    {
      "epoch": 2.0,
      "grad_norm": 11.940149307250977,
      "learning_rate": 4.800378787878788e-05,
      "loss": 0.2393,
      "step": 528
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8337155963302753,
      "eval_loss": 0.40144169330596924,
      "eval_runtime": 0.4501,
      "eval_samples_per_second": 1937.497,
      "eval_steps_per_second": 8.888,
      "step": 528
    },
    {
      "epoch": 3.0,
      "grad_norm": 6.662248611450195,
      "learning_rate": 4.7003787878787884e-05,
      "loss": 0.179,
      "step": 792
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8176605504587156,
      "eval_loss": 0.4737458825111389,
      "eval_runtime": 0.4398,
      "eval_samples_per_second": 1982.627,
      "eval_steps_per_second": 9.095,
      "step": 792
    },
    {
      "epoch": 4.0,
      "grad_norm": 4.980842113494873,
      "learning_rate": 4.600378787878788e-05,
      "loss": 0.1433,
      "step": 1056
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.8348623853211009,
      "eval_loss": 0.44917038083076477,
      "eval_runtime": 0.4416,
      "eval_samples_per_second": 1974.571,
      "eval_steps_per_second": 9.058,
      "step": 1056
    },
    {
      "epoch": 5.0,
      "grad_norm": 4.454750061035156,
      "learning_rate": 4.5003787878787876e-05,
      "loss": 0.1139,
      "step": 1320
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.8325688073394495,
      "eval_loss": 0.492689311504364,
      "eval_runtime": 0.4416,
      "eval_samples_per_second": 1974.846,
      "eval_steps_per_second": 9.059,
      "step": 1320
    },
    {
      "epoch": 6.0,
      "grad_norm": 9.002946853637695,
      "learning_rate": 4.4003787878787885e-05,
      "loss": 0.0936,
      "step": 1584
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.8532110091743119,
      "eval_loss": 0.4423736035823822,
      "eval_runtime": 0.4481,
      "eval_samples_per_second": 1946.081,
      "eval_steps_per_second": 8.927,
      "step": 1584
    },
    {
      "epoch": 7.0,
      "grad_norm": 1.692452073097229,
      "learning_rate": 4.300378787878788e-05,
      "loss": 0.0778,
      "step": 1848
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.8348623853211009,
      "eval_loss": 0.5840200781822205,
      "eval_runtime": 0.4328,
      "eval_samples_per_second": 2014.892,
      "eval_steps_per_second": 9.243,
      "step": 1848
    },
    {
      "epoch": 7.0,
      "step": 1848,
      "total_flos": 3.1225413912290304e+16,
      "train_loss": 0.17279978954430783,
      "train_runtime": 463.2148,
      "train_samples_per_second": 7269.738,
      "train_steps_per_second": 28.496
    }
  ],
  "logging_steps": 1,
  "max_steps": 13200,
  "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": 3.1225413912290304e+16,
  "train_batch_size": 256,
  "trial_name": null,
  "trial_params": null
}