Elvir Crnčević comments

Repositories
Issues
Comments

Results 4 comments of


                                            Elvir Crnčević

Unexpected Performance Regression

Whoops, seems like this code path doesn't invoke mma instructions (please correct me if I'm wrong). I guess I could get away with just having multiple m16n8k16, but I am...

forward function must have ground transform

Have you tried doing this: https://github.com/qinzheng93/GeoTransformer/issues/116?

Why NUM_MAX_NVL_PEERS must be 8?

@LyricZhao Hitting this assert when switching NUM_MAX_NVL_PEERS to 4: ``` DeepEP/csrc/kernels/internode.cu(295): error: static assertion failed with "Invalid number of NVL peers" static_assert(4 * sizeof(bool) == sizeof(uint64_t), "Invalid number of NVL...

Can deep_ep run on environments with more than 160 ranks?

@goelayu Did you modify this https://github.com/fzyzcjy/DeepEP/blob/483f00af8490b0cc378823c6adecf9ea67602071/csrc/kernels/launch.cuh#L54 to scale up the ranks?