checkpoint-500/trainer_state.json

{
  "best_metric": 1.2226382493972778,
  "best_model_checkpoint": "./output/clip-finetuned-csu-p14-336-e3l37-l/checkpoint-500",
  "epoch": 0.056287290329843524,
  "eval_steps": 500,
  "global_step": 500,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.056287290329843524,
      "grad_norm": 96.92656707763672,
      "learning_rate": 2.9437127096701565e-07,
      "loss": 0.3748,
      "step": 500
    },
    {
      "epoch": 0.056287290329843524,
      "eval_loss": 1.2226382493972778,
      "eval_runtime": 131.2779,
      "eval_samples_per_second": 15.037,
      "eval_steps_per_second": 1.882,
      "step": 500
    }
  ],
  "logging_steps": 500,
  "max_steps": 26649,
  "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": 180598692332520.0,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}