[optional format]

Identify the file to be fixed

Describe the problem
vllm serve gpt-oss command will download all the files from huggingface repo, but i think metal/model.bin is not required to run the model with vllm