checkpoint-4000/trainer_state.json

{
  "best_metric": 5.884151935577393,
  "best_model_checkpoint": "/home/p318482/babyLM_controlled/models_trained/de_clm/childes_30/checkpoint-4000",
  "epoch": 4.1928721174004195,
  "eval_steps": 2000,
  "global_step": 4000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.0964360587002098,
      "eval_loss": 7.102903366088867,
      "eval_runtime": 0.9708,
      "eval_samples_per_second": 1416.286,
      "eval_steps_per_second": 88.582,
      "step": 2000
    },
    {
      "epoch": 4.1928721174004195,
      "grad_norm": 1.3964662551879883,
      "learning_rate": 1e-05,
      "loss": 6.9987,
      "step": 4000
    },
    {
      "epoch": 4.1928721174004195,
      "eval_loss": 5.884151935577393,
      "eval_runtime": 0.966,
      "eval_samples_per_second": 1423.408,
      "eval_steps_per_second": 89.028,
      "step": 4000
    }
  ],
  "logging_steps": 4000,
  "max_steps": 100000,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 105,
  "save_steps": 4000,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 1034264024727552.0,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}