trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.7619047619047619,
  "eval_steps": 500,
  "global_step": 2,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "clip_ratio": 0.0,
      "completion_length": 760.5469055175781,
      "epoch": 0.38095238095238093,
      "grad_norm": 0.19004827737808228,
      "kl": 0.0,
      "learning_rate": 3e-06,
      "loss": 0.0244,
      "reward": 0.2343750111758709,
      "reward_std": 0.2009137775748968,
      "rewards/accuracy_reward": 0.19754465017467737,
      "rewards/format_reward": 0.036830359254963696,
      "step": 1
    },
    {
      "clip_ratio": 0.0,
      "completion_length": 749.5469131469727,
      "epoch": 0.7619047619047619,
      "kl": 0.0,
      "reward": 0.20424108020961285,
      "reward_std": 0.26778374426066875,
      "rewards/accuracy_reward": 0.17410715110599995,
      "rewards/format_reward": 0.030133930151350796,
      "step": 2,
      "total_flos": 0.0,
      "train_loss": 0.027519498951733112,
      "train_runtime": 646.1864,
      "train_samples_per_second": 0.52,
      "train_steps_per_second": 0.003
    }
  ],
  "logging_steps": 5,
  "max_steps": 2,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": false,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}