train_simple.py [91:101]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        },
    ),
    ModelConfig(
        name="Qwen/Qwen-72B",
        default_lr=1e-5,
        eval_batch_size=1,
        gradient_checkpointing=True,
        model_parallel=True,
        # note: you will probably not be able to run this without bf16 support and many gpus
        custom_kwargs={
            "trust_remote_code": True,
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



train_weak_to_strong.py [85:95]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        },
    ),
    ModelConfig(
        name="Qwen/Qwen-72B",
        default_lr=1e-5,
        eval_batch_size=1,
        gradient_checkpointing=True,
        model_parallel=True,
        # note: you will probably not be able to run this without bf16 support and many gpus
        custom_kwargs={
            "trust_remote_code": True,
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -