-
Notifications
You must be signed in to change notification settings - Fork 2.2k
Pull requests: vllm-project/vllm-ascend
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[CI] Improve select node
module:tests
ready-all
run all e2e test for pr
#15595
opened Sep 2, 2026 by
wangxiyuan
Collaborator
Loading…
[BugFix][Mamba] Overlap-safe state copies for align-mode block moves
module:ops
#15592
opened Sep 2, 2026 by
Liears
Loading…
[Bugfix][KVPool] Cover physical DCP shards in scheduler lookup
module:tests
ready-all
run all e2e test for pr
#15591
opened Sep 2, 2026 by
zmc1997
Contributor
Loading…
[Ops] Cache packed GDN conv1d weight instead of repacking each forward
module:ops
module:tests
#15590
opened Sep 2, 2026 by
leijie-ww
Contributor
Loading…
[Ops] Split fused qkvz projection into separate qkv and z linears in GDN
module:ops
module:tests
#15589
opened Sep 2, 2026 by
leijie-ww
Contributor
Loading…
perf: fuse MLA query projection transposes
module:tests
#15588
opened Sep 2, 2026 by
sustech-lz
Loading…
[Performance][Attention] Fuse MLA query projection transposes
ci/build
documentation
Improvements or additions to documentation
merge-conflicts
module:core
module:ops
module:quantization
module:tests
module:tools
#15587
opened Sep 2, 2026 by
sustech-lz
Loading…
[Bugfix] Support standardized KV-cache layouts on current vLLM main
module:tests
#15585
opened Sep 2, 2026 by
ShuhaoZhangTony
Loading…
[Performance][Worker] Build DeepSeek V4 metadata asynchronously on v0.26
module:core
module:tests
#15584
opened Sep 2, 2026 by
freyfwt
Collaborator
Loading…
[Ops][Performance] Warm up GDN prefill kernels before KV-cache allocation
module:tests
#15583
opened Sep 2, 2026 by
prefersung
•
Draft
[Feature][SpecDecode] Backport request-aligned DSpark ACLGraph to v0.26
#15582
opened Sep 2, 2026 by
luo-rabbit
Loading…
[CI] enable cpu test parallel with pre-commit
ci/build
#15580
opened Sep 2, 2026 by
xqchen7
Contributor
Loading…
perf(attention): reuse PA workspace across layers in graph param update
#15579
opened Sep 2, 2026 by
solin28825
Loading…
Remove unused npu_flash_attention wrappers and their UTs
module:tests
#15576
opened Sep 2, 2026 by
ZT-AIA
Collaborator
Loading…
[Ops][Feature] Enable Qwen4Exp (Qwen3.8-Flash-Next) on Ascend NPU
module:tests
#15574
opened Sep 2, 2026 by
krell11
Loading…
[Feature][MRV2] Optimize resample_kernel with 150x speed up
module:ops
#15573
opened Sep 2, 2026 by
AuroraEmiya
Contributor
Loading…
[Feature][Ops] Add fused FFN linear op (ffn_linear) for Ascend950
documentation
Improvements or additions to documentation
module:tests
#15572
opened Sep 2, 2026 by
chenchris2
Loading…
[Feature][Multimodal] Enable encoder acl graph for model runner v2
module:ops
module:tests
#15571
opened Sep 2, 2026 by
HF-001
Collaborator
Loading…
[Feature][KV Transfer] Support multiple memfabric TransDataOpType and A5
documentation
Improvements or additions to documentation
module:tests
#15570
opened Sep 2, 2026 by
dsxsteven
Contributor
Loading…
[Performance][Attention] Fuse local SFA LSE formation into DCP pack
module:ops
module:tests
#15568
opened Sep 2, 2026 by
Ruiqiu-Zheng
Loading…
[CI]Recitify Qwen3.8-27B spec decode E2E
module:tests
#15567
opened Sep 2, 2026 by
sunny-rain-63
Loading…
[Feature][Quantization]Support NZ for A5 MX quantization
module:core
module:ops
module:quantization
module:tests
#15565
opened Sep 2, 2026 by
lijiahang226
Collaborator
Loading…
Revert "[Feature] GLM5.2 rotation v2 Dspark support"
merge-conflicts
module:core
module:tests
#15564
opened Sep 2, 2026 by
Tflowers-0129
Collaborator
Loading…
[CI]: collect and upload test case metadata and JUnit results
ci/build
ready-a5
#15559
opened Sep 2, 2026 by
zhangxinyuehfad
Collaborator
Loading…
fix(mla): drop decode head padding for non-power-of-two heads
module:tests
#15558
opened Sep 2, 2026 by
MaybeChz
Contributor
Loading…
Previous Next
ProTip!
Find all pull requests that aren't related to any open issues with -linked:issue.