Skip to content

Pull requests: xLLM-AI/xllm

New pull request New
97 Open 1,648 Closed
97 Open 1,648 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

bugfix: defer stream cancellation to scheduler boundary.
#1926 opened Jul 11, 2026 by phantomlei3 Collaborator Loading…
9 of 17 tasks
bugfix: set up kv cache compleition guard to prevent kv cache transfer error.
#1925 opened Jul 11, 2026 by phantomlei3 Collaborator Loading…
9 of 17 tasks
perf: avoid mega chunk sequence length device reads on npu.
#1924 opened Jul 11, 2026 by Sinle4Cat Contributor Loading…
feat: add NPU fused norm operators for Wan2.2 DiT.
#1923 opened Jul 10, 2026 by ethan686 Collaborator Loading…
feat: add speculative decoding per-token latency metrics.
#1922 opened Jul 10, 2026 by phantomlei3 Collaborator Loading…
10 of 17 tasks
feat: add musa_graph support metadata.
#1920 opened Jul 10, 2026 by fay85 Collaborator Loading…
8 of 17 tasks
feat(npu): route gated norm to aclnn layer norm fwd
#1919 opened Jul 10, 2026 by Fengfengst123 • Draft
1
feat: add RWKV-7-World model support on cuda device.
#1918 opened Jul 10, 2026 by Dragonliu2018 Collaborator Loading…
9 of 17 tasks
2
feat: add qwen3.5 dcu support.
#1917 opened Jul 10, 2026 by cding-nv Collaborator Loading…
8 of 17 tasks
feat: support deepseek-v3 fp8 w8a8 on dcu device.
#1915 opened Jul 9, 2026 by asr-sheep1 Collaborator Loading…
8 of 17 tasks
feat: support Wan2.2 distill model.
#1910 opened Jul 8, 2026 by ethan686 Collaborator Loading…
1
feat: add request trace dump with runtime toggle.
#1906 opened Jul 7, 2026 by shifengmin Collaborator Loading…
8 of 17 tasks
3
feat: implement single-process mode on npu device.
#1894 opened Jul 6, 2026 by XuZhang99 Collaborator Loading…
8 of 17 tasks
5
feat: add embedded-python model executor for qwen3 on cuda.
#1891 opened Jul 6, 2026 by zhang-minchao Collaborator Loading…
5 of 7 tasks
78
feat: Add FlashComm1 and MMRS fusion support for Qwen3.5 NPU.
#1890 opened Jul 6, 2026 by Cooofish Loading…
21
perf: route Qwen3.5 FSGDR model path to TileLang.
#1889 opened Jul 6, 2026 by ShareableXue Contributor Loading…
7 of 17 tasks
2
feat: add DFlash block-diffusion speculative decoding on npu device.
#1886 opened Jul 3, 2026 by ustcfy Collaborator Loading…
2 of 7 tasks
22
feat: support adaptive speculative decoding.
#1876 opened Jul 2, 2026 by weizhehuang0827 Collaborator Loading…
7 of 17 tasks
10
Cp chunkprefill upstream 20260610 latest
#1872 opened Jul 2, 2026 by cchh05 Collaborator Loading…
17 tasks
6
feat: add ViT encoder ACL Graph capture for Qwen3-VL.
#1866 opened Jul 1, 2026 by sunbaosong Collaborator Loading…
19
docs: remove docs dir.
#1865 opened Jul 1, 2026 by XuZhang99 Collaborator Loading…
7 of 17 tasks
1
feat: support multi machine for offline inference.
#1863 opened Jul 1, 2026 by weizhehuang0827 Collaborator Loading…
9 of 12 tasks
2
Ref blockmgr4
#1861 opened Jul 1, 2026 by Kang-Meng Collaborator • Draft
17 tasks
10
feat: add prefix-cache-affinity dp rank routing.
#1860 opened Jul 1, 2026 by shifengmin Collaborator Loading…
7 of 15 tasks
6
feat: enable SwiGluQuant fusion for Qwen3 MLP
#1847 opened Jun 29, 2026 by nie-linfeng Contributor Loading…
4
ProTip! Adding no:label will show everything without a label.