checkpoint-625/trainer_state.json

{
  "best_global_step": 625,
  "best_metric": 0.3823908269405365,
  "best_model_checkpoint": "siglip2-finetune-full/checkpoint-625",
  "epoch": 1.0,
  "eval_steps": 500,
  "global_step": 625,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.8,
      "grad_norm": 8.966306686401367,
      "learning_rate": 1.63265306122449e-05,
      "loss": 0.5146,
      "step": 500
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.80765,
      "eval_loss": 0.3823908269405365,
      "eval_model_preparation_time": 0.0027,
      "eval_runtime": 411.2751,
      "eval_samples_per_second": 48.629,
      "eval_steps_per_second": 6.079,
      "step": 625
    }
  ],
  "logging_steps": 500,
  "max_steps": 2500,
  "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": 1.67510991642624e+18,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": null
}