Skip to content

BugFix for MLAPO for Deepseek eagle3 on Ascend #222

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
realray808 merged 1 commit into from
Apr 3, 2026
+2 −2
Merged

BugFix for MLAPO for Deepseek eagle3 on Ascend #222

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions python/sglang/srt/hardware_backend/npu/attention/ascend_backend.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -840,7 +840,7 @@ def forward_extend(
sinks: Optional[torch.Tensor] = None,
slopes: Optional[torch.Tensor] = None,
):
if is_mla_preprocess_enabled():
if is_mla_preprocess_enabled() and self.use_mla:
# MLAPO and MLAPROLOG do save kv_cache
save_kv_cache = False
if self.is_dllm_model:
Expand Down Expand Up @@ -1748,7 +1748,7 @@ def forward_decode(
sinks: Optional[torch.Tensor] = None,
slopes: Optional[torch.Tensor] = None,
):
if is_mla_preprocess_enabled():
if is_mla_preprocess_enabled() and self.use_mla:
# MLAPO does saving kv_cache
save_kv_cache = False
if topk_indices is not None:
Expand Down
Loading