[CI/Build] Per file CUDA Archs (improve wheel size and dev build times) (#8845)
This commit is contained in:
@@ -1,6 +1,5 @@
|
||||
#include "core/registration.h"
|
||||
#include "moe_ops.h"
|
||||
#include "marlin_moe_ops.h"
|
||||
|
||||
TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, m) {
|
||||
// Apply topk softmax to the gating outputs.
|
||||
@@ -18,7 +17,7 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, m) {
|
||||
"int size_n, int size_k, bool is_k_full, int num_experts, int topk, "
|
||||
"int moe_block_size, bool replicate_input, bool apply_weights)"
|
||||
" -> Tensor");
|
||||
m.impl("marlin_gemm_moe", torch::kCUDA, &marlin_gemm_moe);
|
||||
// conditionally compiled so impl registration is in source file
|
||||
#endif
|
||||
}
|
||||
|
||||
|
||||
Reference in New Issue
Block a user