Fix amd rope definition (#14556)

This commit is contained in:
Qiaolin Yu
2025-12-07 23:47:03 -08:00
committed by GitHub
parent cf0478d602
commit aeff0d386b
2 changed files with 10 additions and 0 deletions
+9
View File
@@ -186,6 +186,15 @@ TORCH_LIBRARY_EXPAND(sgl_kernel, m) {
*/
m.def("apply_token_bitmask_inplace_cuda(Tensor logits, Tensor bitmask, Tensor? indices=None) -> ()");
m.impl("apply_token_bitmask_inplace_cuda", &ApplyTokenBitmaskInplace);
/*
* From csrc/elementwise
*/
m.def(
"rotary_embedding(Tensor positions, Tensor! query,"
" Tensor!? key, int head_size,"
" Tensor cos_sin_cache, bool is_neox) -> ()");
m.impl("rotary_embedding", torch::kCUDA, &rotary_embedding);
}
REGISTER_EXTENSION(common_ops)