checkpoint-1021/trainer_state.json

{
  "best_global_step": 1021,
  "best_metric": 1.7632426023483276,
  "best_model_checkpoint": "siglip2-finetune-full/checkpoint-1021",
  "epoch": 1.0,
  "eval_steps": 500,
  "global_step": 1021,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.48971596474045054,
      "grad_norm": 59.27558135986328,
      "learning_rate": 2.665344571145265e-06,
      "loss": 4.268,
      "step": 500
    },
    {
      "epoch": 0.9794319294809011,
      "grad_norm": 61.66608810424805,
      "learning_rate": 2.2935052057511156e-06,
      "loss": 2.4217,
      "step": 1000
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.666574839302112,
      "eval_loss": 1.7632426023483276,
      "eval_model_preparation_time": 0.0024,
      "eval_runtime": 476.9192,
      "eval_samples_per_second": 68.502,
      "eval_steps_per_second": 8.563,
      "step": 1021
    }
  ],
  "logging_steps": 500,
  "max_steps": 4084,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 2.738992217671987e+18,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": null
}