checkpoint-20/trainer_state.json

{
  "best_metric": 0.467332661151886,
  "best_model_checkpoint": "./mistral/01-03-24-Weni-ZeroShot-3.3.18-Mistral-7b-Multilanguage-3.2.0_Zeroshot-2_max_steps-100_batch_16_2024-03-01_ppid_7/checkpoint-20",
  "epoch": 0.19801980198019803,
  "eval_steps": 10,
  "global_step": 20,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.1,
      "eval_loss": 0.6981944441795349,
      "eval_runtime": 94.6462,
      "eval_samples_per_second": 30.292,
      "eval_steps_per_second": 0.951,
      "step": 10
    },
    {
      "epoch": 0.2,
      "grad_norm": 0.2654201090335846,
      "learning_rate": 0.0001961261695938319,
      "loss": 1.0016,
      "step": 20
    },
    {
      "epoch": 0.2,
      "eval_loss": 0.467332661151886,
      "eval_runtime": 94.5373,
      "eval_samples_per_second": 30.327,
      "eval_steps_per_second": 0.952,
      "step": 20
    }
  ],
  "logging_steps": 20,
  "max_steps": 100,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 10,
  "total_flos": 1.7984776120041472e+17,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}