checkpoint-2380/trainer_state.json

{
  "best_global_step": 2380,
  "best_metric": 0.23066656291484833,
  "best_model_checkpoint": "siglip2-finetune-full/checkpoint-2380",
  "epoch": 4.0,
  "eval_steps": 500,
  "global_step": 2380,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.8403361344537815,
      "grad_norm": 37.12611770629883,
      "learning_rate": 4.034334763948498e-06,
      "loss": 0.757,
      "step": 500
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8265671013883046,
      "eval_loss": 0.44112637639045715,
      "eval_model_preparation_time": 0.0026,
      "eval_runtime": 308.0267,
      "eval_samples_per_second": 61.735,
      "eval_steps_per_second": 7.717,
      "step": 595
    },
    {
      "epoch": 1.680672268907563,
      "grad_norm": 40.48983383178711,
      "learning_rate": 2.96137339055794e-06,
      "loss": 0.4737,
      "step": 1000
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8392406394615061,
      "eval_loss": 0.41642823815345764,
      "eval_model_preparation_time": 0.0026,
      "eval_runtime": 305.306,
      "eval_samples_per_second": 62.285,
      "eval_steps_per_second": 7.786,
      "step": 1190
    },
    {
      "epoch": 2.5210084033613445,
      "grad_norm": 30.135860443115234,
      "learning_rate": 1.8884120171673823e-06,
      "loss": 0.3811,
      "step": 1500
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.889724442574674,
      "eval_loss": 0.2768375873565674,
      "eval_model_preparation_time": 0.0026,
      "eval_runtime": 304.3734,
      "eval_samples_per_second": 62.476,
      "eval_steps_per_second": 7.809,
      "step": 1785
    },
    {
      "epoch": 3.361344537815126,
      "grad_norm": 35.56352233886719,
      "learning_rate": 8.15450643776824e-07,
      "loss": 0.2957,
      "step": 2000
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.9109171224232225,
      "eval_loss": 0.23066656291484833,
      "eval_model_preparation_time": 0.0026,
      "eval_runtime": 300.2689,
      "eval_samples_per_second": 63.33,
      "eval_steps_per_second": 7.916,
      "step": 2380
    }
  ],
  "logging_steps": 500,
  "max_steps": 2380,
  "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": true
      },
      "attributes": {}
    }
  },
  "total_flos": 6.370936521755001e+18,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": null
}