hclearner
hclearner
Do you have any plans, such as a thin wrapper library?
Hi @bartekxk ，the two implementations are the same, the second version is more preferred for me. I think this is just the layout using tensor descriptor in ck, it is...
@bartekxk Cool！We're hipifying the library like xformer in AMD GPU. The operation is mem_efficient_attention. https://github.com/facebookresearch/xformers/tree/main/xformers/csrc/attention/cuda/fmha
@ > Hi @hclearner , thanks. Maybe I'm wrong but I don't see CuTe usage in https://github.com/facebookresearch/xformers/tree/main/xformers/csrc/attention/cuda/fmha. Or do you need this just because it would be easier for You...
Do you have any plans for `cutlass`？ thanks :) I think this is very import.