[Kernel][Misc] Use TORCH_LIBRARY instead of PYBIND11_MODULE for custom ops (#5047)

This commit is contained in:
bnellnm
2024-06-09 16:23:30 -04:00
committed by GitHub
parent 5d7e3d0176
commit 5467ac3196
55 changed files with 833 additions and 451 deletions
+12
View File
@@ -0,0 +1,12 @@
#include "registration.h"
#include "moe_ops.h"
TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, m) {
// Apply topk softmax to the gating outputs.
m.def(
"topk_softmax(Tensor! topk_weights, Tensor! topk_indices, Tensor! "
"token_expert_indices, Tensor gating_output) -> ()");
m.impl("topk_softmax", torch::kCUDA, &topk_softmax);
}
REGISTER_EXTENSION(TORCH_EXTENSION_NAME)