Does QQQ linear support H100?

[donglinz]

I tried to run it on H100, but it seems there is an illegal memory access inside the kernel.

RuntimeError: CUDA error: an illegal memory access was encountered

[HandH1998]

We don't have a H100, and we don't try it on H100. But the vllm viewer seems to have tried it on H100 in vllm-project/vllm#5218 (comment). Our PR has been merged into vllm now, you can try qqq in vllm on H100 and see if it can work.