trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.8,
  "eval_steps": 500,
  "global_step": 20,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "clip_ratio": 0.0,
      "completion_length": 994.5480270385742,
      "epoch": 0.2,
      "grad_norm": 150.32078552246094,
      "kl": 0.0,
      "learning_rate": 1e-06,
      "loss": -0.0199,
      "reward": 0.7332589626312256,
      "reward_std": 0.41786693036556244,
      "rewards/accuracy_reward": 0.06808035937137902,
      "rewards/format_reward": 0.6651786044239998,
      "step": 1
    },
    {
      "clip_ratio": 0.0,
      "completion_length": 992.7260551452637,
      "epoch": 1.2,
      "grad_norm": 27.00591468811035,
      "kl": 0.785003662109375,
      "learning_rate": 2.9392394604217463e-06,
      "loss": 0.0082,
      "reward": 0.7229353059083223,
      "reward_std": 0.3804802359081805,
      "rewards/accuracy_reward": 0.049386162601877004,
      "rewards/format_reward": 0.673549136146903,
      "step": 5
    },
    {
      "clip_ratio": 0.0,
      "completion_length": 1002.7217468261719,
      "epoch": 2.4,
      "grad_norm": 0.172083780169487,
      "kl": 2.3208984375,
      "learning_rate": 2.3109612261833968e-06,
      "loss": 0.0779,
      "reward": 0.7736607521772385,
      "reward_std": 0.3761067040264606,
      "rewards/accuracy_reward": 0.053348216926679015,
      "rewards/format_reward": 0.7203125387430191,
      "step": 10
    },
    {
      "clip_ratio": 0.0,
      "completion_length": 1003.5186981201172,
      "epoch": 3.6,
      "grad_norm": 0.11088093370199203,
      "kl": 2.690625,
      "learning_rate": 1.2865277425900725e-06,
      "loss": 0.104,
      "reward": 0.8734375417232514,
      "reward_std": 0.3127763744443655,
      "rewards/accuracy_reward": 0.05937500293366611,
      "rewards/format_reward": 0.8140625327825546,
      "step": 15
    },
    {
      "clip_ratio": 0.0,
      "completion_length": 998.7733978271484,
      "epoch": 4.8,
      "grad_norm": 0.18645578622817993,
      "kl": 2.891796875,
      "learning_rate": 3.6637563846861275e-07,
      "loss": 0.1122,
      "reward": 0.9082589730620384,
      "reward_std": 0.2693608712404966,
      "rewards/accuracy_reward": 0.058258931757882236,
      "rewards/format_reward": 0.8500000417232514,
      "step": 20
    },
    {
      "epoch": 4.8,
      "step": 20,
      "total_flos": 0.0,
      "train_loss": 0.07416041865944863,
      "train_runtime": 6285.2191,
      "train_samples_per_second": 0.502,
      "train_steps_per_second": 0.004
    }
  ],
  "logging_steps": 5,
  "max_steps": 25,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "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
}