checkpoint-1250/trainer_state.json

{
  "best_global_step": 1250,
  "best_metric": 0.33390238881111145,
  "best_model_checkpoint": "siglip2-finetune-full/checkpoint-1250",
  "epoch": 2.0,
  "eval_steps": 500,
  "global_step": 1250,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.8,
      "grad_norm": 48.33537673950195,
      "learning_rate": 1.5068493150684933e-05,
      "loss": 0.5378,
      "step": 500
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7981,
      "eval_loss": 0.40214166045188904,
      "eval_model_preparation_time": 0.0025,
      "eval_runtime": 436.5615,
      "eval_samples_per_second": 45.813,
      "eval_steps_per_second": 5.727,
      "step": 625
    },
    {
      "epoch": 1.6,
      "grad_norm": 28.126569747924805,
      "learning_rate": 9.589041095890411e-06,
      "loss": 0.4136,
      "step": 1000
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.84045,
      "eval_loss": 0.33390238881111145,
      "eval_model_preparation_time": 0.0025,
      "eval_runtime": 440.1327,
      "eval_samples_per_second": 45.441,
      "eval_steps_per_second": 5.68,
      "step": 1250
    }
  ],
  "logging_steps": 500,
  "max_steps": 1875,
  "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": 3.35021983285248e+18,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": null
}