biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Refactor] Remove align block size logic in moe_permute (#33449)

Browse Source 
Signed-off-by: yewentao256 <zhyanwentao@126.com>
This commit is contained in:
Wentao Ye
2026-02-06 13:57:06 -05:00
committed by GitHub
parent 16786da735
commit 77c09e1130
8 changed files with 38 additions and 297 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
csrc/moe/torch_bindings.cpp
View File

@@ -99,9 +99,9 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, m) {
"moe_permute(Tensor input, Tensor topk_ids,"
"Tensor token_expert_indices, Tensor? expert_map, int n_expert,"
"int n_local_expert,"
"int topk, int? align_block_size,Tensor! permuted_input, Tensor! "
"int topk, Tensor! permuted_input, Tensor! "
"expert_first_token_offset, Tensor! inv_permuted_idx, Tensor! "
"permuted_idx, Tensor! m_indices)->()");
"permuted_idx)->()");
m.def(
"moe_unpermute(Tensor permuted_hidden_states, Tensor topk_weights,"