Skip to content

Pull requests: vllm-project/vllm-ascend

New pull request New
154 Open 1,795 Closed
154 Open 1,795 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

replace npu_incre_flash_attention with npu_fused_infer_attention_score
#2792 opened Sep 6, 2025 by panchao-hub Loading…
3
refactor fused_moe.py module:core module:ops module:quantization module:tests
#2791 opened Sep 6, 2025 by Pr0Wh1teGivee Loading…
3
Deepseek Mtp model uses the lm_head and embedding from the main model module:tests
#2790 opened Sep 5, 2025 by zzhx1 Loading…
5
[Disagg][Perf] Use NPU event sync instead of blocking tolist to avoid unintentional copy ops blocking across different NPU streams, improving disagg TTIT/TTFT module:tests
#2788 opened Sep 5, 2025 by jesse996 Loading…
4
fix qwen torchair attention PrefillCacheHit
#2787 opened Sep 5, 2025 by zhaozx-cn Loading…
4
[WIP] Fix the bugs about operator registration by PyTorch Dispatcher module:ops module:tests
#2786 opened Sep 5, 2025 by FFFrog • Draft
8
[2/N][Refactor][Quantization] clean quantization patch module:ops module:quantization module:tests
#2785 opened Sep 5, 2025 by 22dimensions Loading…
3
remove chunked prefill and fix ring mla mask
#2784 opened Sep 5, 2025 by SunnyLee151064 Loading…
2
[Perf][V1] Fully overlap model execution merge-conflicts
#2783 opened Sep 5, 2025 by jiangpeng36 Loading…
7
[Benchmark] Correctly kill vllm process in performance benchamrk performance-test enable performance test for PR ready-for-test start test by label for PR
#2782 opened Sep 5, 2025 by Potabk Loading…
3
Remove chunked_prefill_for_mla and fix ring_mla bug documentation Improvements or additions to documentation module:core
#2781 opened Sep 5, 2025 by SunnyLee151064 Loading…
4
support qwen25 vl w8a8 quantization module:quantization module:tests
#2778 opened Sep 5, 2025 by wenba0 Loading…
5
support qwen25 vl w8a8 quantization module:quantization module:tests
#2777 opened Sep 5, 2025 by wenba0 Loading…
5
Refactor the Spec decode module to merge MTP non-torchair and eagle modes into one file, separating torchair and non-torchair modes #2773
#2776 opened Sep 5, 2025 by weisirui-eng Loading…
3
[0.9.1][Doc] Update supported models documentation Improvements or additions to documentation
#2774 opened Sep 5, 2025 by zhangxinyuehfad Loading…
7
[main] addrmsnorm + quant fusion optim in Qwen Models module:core module:ops
#2772 opened Sep 5, 2025 by rjg-lyh Loading…
5
Install vllm from source to make doctest passed documentation Improvements or additions to documentation module:tests
#2769 opened Sep 5, 2025 by Yikun • Draft
2
[Fix] Ensure metadata sync across DP ranks in eager mode
#2766 opened Sep 5, 2025 by yiz-liu Loading…
4
Add note for Ascend HDK version documentation Improvements or additions to documentation
#2765 opened Sep 5, 2025 by Yikun Loading…
1
[main] mlp weight prefetch in Qwen Dense Models module:core module:ops module:tests
#2762 opened Sep 4, 2025 by rjg-lyh Loading…
6
[fix] prefill unsupport sliding window attention module:tests
#2758 opened Sep 4, 2025 by NSDie Loading…
7
add weight transpose check. module:ops
#2756 opened Sep 4, 2025 by momo609 Loading…
6
[main] add pd transfer for ascend scheduler module:ops module:tests
#2753 opened Sep 4, 2025 by Liccol Loading…
6
[Feat] communication optimization for mc2 ops on A2 module:ops module:tests
#2752 opened Sep 4, 2025 by realliujiaxu Loading…
5
[DOC] Qwen3 PD disaggregation user guide documentation Improvements or additions to documentation ready read for review
#2751 opened Sep 4, 2025 by paulyu12 Loading…
13
ProTip! Mix and match filters to narrow down what you’re looking for.