checkpoint-30/trainer_state.json

{
  "best_metric": 0.44581571221351624,
  "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-30",
  "epoch": 0.14869888475836432,
  "eval_steps": 10,
  "global_step": 30,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.05,
      "eval_loss": 0.6367942690849304,
      "eval_runtime": 180.8957,
      "eval_samples_per_second": 15.849,
      "eval_steps_per_second": 0.995,
      "step": 10
    },
    {
      "epoch": 0.1,
      "grad_norm": 0.34346863627433777,
      "learning_rate": 0.00019510565162951537,
      "loss": 0.9774,
      "step": 20
    },
    {
      "epoch": 0.1,
      "eval_loss": 0.48089343309402466,
      "eval_runtime": 180.8882,
      "eval_samples_per_second": 15.85,
      "eval_steps_per_second": 0.995,
      "step": 20
    },
    {
      "epoch": 0.15,
      "eval_loss": 0.44581571221351624,
      "eval_runtime": 180.8305,
      "eval_samples_per_second": 15.855,
      "eval_steps_per_second": 0.995,
      "step": 30
    }
  ],
  "logging_steps": 20,
  "max_steps": 100,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 10,
  "total_flos": 1.3497349192024064e+17,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}