trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.8,
  "eval_steps": 500,
  "global_step": 4,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "clip_ratio": 0.0,
      "completion_length": 794.7712326049805,
      "epoch": 0.2,
      "grad_norm": 0.1649693101644516,
      "kl": 0.0,
      "learning_rate": 3e-06,
      "loss": 0.0129,
      "reward": 0.1897321529686451,
      "reward_std": 0.19452152773737907,
      "rewards/accuracy_reward": 0.16183036379516125,
      "rewards/format_reward": 0.027901787078008056,
      "step": 1
    },
    {
      "clip_ratio": 0.0,
      "completion_length": 789.3489176432291,
      "epoch": 0.8,
      "kl": 0.0007623036702473959,
      "reward": 0.16108631622046232,
      "reward_std": 0.20495034692188105,
      "rewards/accuracy_reward": 0.12537203108270964,
      "rewards/format_reward": 0.03571428781530509,
      "step": 4,
      "total_flos": 0.0,
      "train_loss": 0.025162240955978632,
      "train_runtime": 1296.9376,
      "train_samples_per_second": 0.487,
      "train_steps_per_second": 0.004
    }
  ],
  "logging_steps": 5,
  "max_steps": 5,
  "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
}