checkpoint-1062/trainer_state.json

{
  "best_global_step": 1062,
  "best_metric": 0.7104782462120056,
  "best_model_checkpoint": "siglip2-finetune-full/checkpoint-1062",
  "epoch": 1.0,
  "eval_steps": 500,
  "global_step": 1062,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.4708097928436911,
      "grad_norm": 10.774321556091309,
      "learning_rate": 8.565051020408164e-05,
      "loss": 1.047,
      "step": 500
    },
    {
      "epoch": 0.9416195856873822,
      "grad_norm": 4.252574920654297,
      "learning_rate": 6.970663265306123e-05,
      "loss": 0.7732,
      "step": 1000
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.6644597457627118,
      "eval_loss": 0.7104782462120056,
      "eval_model_preparation_time": 0.0038,
      "eval_runtime": 419.7148,
      "eval_samples_per_second": 80.969,
      "eval_steps_per_second": 10.121,
      "step": 1062
    }
  ],
  "logging_steps": 500,
  "max_steps": 3186,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "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": 2.846393976227365e+18,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": null
}