RobertLiu0905

I grabbed the flame chart, and the problem was gptq.py#apply_weights#ops.gptq_gemm。May be beyond the capacity of cuda computing。  

I also encountered this problem. I solved it by compiling the [nccl source code](https://github.com/NVIDIA/nccl) and then modifying the path of libnccl.so.2 in the vllm source code.

Expect the same