minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

Remove obsolete sgl-kernel legacy paths (#21528)

Browse Source
This commit is contained in:
Xiaoyu Zhang
2026-04-01 09:00:20 +08:00
committed by GitHub
parent a8759dd9af
commit cdd7d6a227
17 changed files with 14 additions and 1708 deletions
Download Patch File Download Diff File Expand all files Collapse all files
sgl-kernel/csrc/common_extension.cc
-10
View File
@@ -50,8 +50,6 @@ TORCH_LIBRARY_FRAGMENT(sgl_kernel, m) {
/*
* From csrc/attention
*/
m.def("merge_state(Tensor v_a, Tensor s_a, Tensor v_b, Tensor s_b, Tensor! v_merged, Tensor! s_merged) -> ()");
m.impl("merge_state", torch::kCUDA, &merge_state);
m.def("merge_state_v2(Tensor v_a, Tensor s_a, Tensor v_b, Tensor s_b, Tensor! v_merged, Tensor! s_merged) -> ()");
m.impl("merge_state_v2", torch::kCUDA, &merge_state_v2);
m.def(
@@ -90,11 +88,6 @@ TORCH_LIBRARY_FRAGMENT(sgl_kernel, m) {
" 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) -> ()");
m.impl("downcast_fp8", torch::kCUDA, &downcast_fp8);
m.def("copy_to_gpu_no_ce(Tensor input, Tensor! output) -> ()");
m.impl("copy_to_gpu_no_ce", torch::kCUDA, &copy_to_gpu_no_ce);
m.def("concat_mla_k(Tensor! k, Tensor k_nope, Tensor k_rope) -> ()");
@@ -364,9 +357,6 @@ TORCH_LIBRARY_FRAGMENT(sgl_kernel, m) {
m.def("top_p_renorm_probs(Tensor probs, Tensor! renorm_probs, Tensor? maybe_top_p_arr, float top_p_val) -> ()");
m.impl("top_p_renorm_probs", torch::kCUDA, &top_p_renorm_probs);
m.def("top_k_mask_logits(Tensor logits, Tensor mask_logits, Tensor? maybe_top_k_arr, int top_k_val) -> ()");
m.impl("top_k_mask_logits", torch::kCUDA, &top_k_mask_logits);
/*
* From Sparse Flash Attention
*/