feng397

I try to run it on H20, and I also encountered the cuda graph error as follows: ``` 2025-05-07 13:37:52 DP7 TP7] DeepGEMM JIT Compiling for M=32, N=7168, K=2048. Please...

> [@feng397](https://github.com/feng397) For that error, could you please try > > [sglang/python/sglang/srt/layers/moe/ep_moe/token_dispatcher.py](https://github.com/sgl-project/sglang/blob/38053c3372dd220911987bd8cb55b27448366497/python/sglang/srt/layers/moe/ep_moe/token_dispatcher.py#L441) > > Line 441 in [38053c3](/sgl-project/sglang/commit/38053c3372dd220911987bd8cb55b27448366497) > > # For H20, there will be an CUDA error: DeepEP/csrc/kernels/internode_ll.cu:337...

I got the same error, what's the version of mooncake of yours?