[1/2] Add rope kernel in sgl-kernel (#14334)
This commit is contained in:
@@ -90,6 +90,12 @@ TORCH_LIBRARY_FRAGMENT(sgl_kernel, m) {
|
||||
"Tensor? v, Tensor!? k_buffer, Tensor!? v_buffer, Tensor? kv_cache_loc) -> ()");
|
||||
m.impl("apply_rope_pos_ids_cos_sin_cache", torch::kCUDA, &apply_rope_pos_ids_cos_sin_cache);
|
||||
|
||||
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);
|
||||
|
||||
m.def(
|
||||
"downcast_fp8(Tensor k, Tensor v, Tensor k_out, Tensor v_out, Tensor k_scale, Tensor v_scale, Tensor loc, "
|
||||
"int mult, int offset) -> ()");
|
||||
|
||||
Reference in New Issue
Block a user