MetaMathQA/experiments/shira/llama-3.2-3B-lr_0.0003-random_seed_42/training_params.json

{
  "optimizer_kwargs": {
    "lr": 3e-4
  }
}