Skip to content

Pull requests: xLLM-AI/xllm

New pull request New
129 Open 1,963 Closed
129 Open 1,963 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

feat: add lag confidence to decouple adaptive speculative pruning from critical path.
#2280 opened Aug 20, 2026 by weizhehuang0827 Collaborator Loading…
4 of 5 tasks
feat: add Qwen3.8 DFlash2 decoding support.
#2279 opened Aug 20, 2026 by pjgao Collaborator • Draft
1
feat: support Qwen3-ASR on NPU device.
#2278 opened Aug 20, 2026 by shan-chen-feng Collaborator Loading…
17 tasks
perf: token-level multimodal mm token expansion.
#2268 opened Aug 19, 2026 by kongweiqian Loading…
9 of 17 tasks
feat: add composite host cache support for MTP.
#2266 opened Aug 19, 2026 by Kang-Meng Collaborator Loading…
9 of 17 tasks
bugfix: make DSV4 SWA capacity PD-aware.
#2265 opened Aug 19, 2026 by Kang-Meng Collaborator Loading…
8 of 17 tasks
feat: support Qwen3.5 eager-mode on NPU with Python executor.
#2264 opened Aug 18, 2026 by Wang-1F Collaborator Loading…
4 of 6 tasks
3
feat: support qwen3 dspark decoding on NPU with Python executor.
#2263 opened Aug 18, 2026 by yinjiawei01 Collaborator Loading…
2
perf: keep raw dit image tensors on cpu until vae preprocess.
#2261 opened Aug 18, 2026 by chuanhaoyu Loading…
8 of 17 tasks
feat: support separate DiT/VAE service instances for Flux.
#2256 opened Aug 17, 2026 by z-jun03 Collaborator • Draft
8 of 17 tasks
feat: add RegionE support for Qwen-Image-Edit.
#2250 opened Aug 17, 2026 by BurnWan Collaborator Loading…
7 of 17 tasks
perf: route low-token qwen3 decode through optimized nz quantmatmul.
#2248 opened Aug 17, 2026 by maojunx99 Collaborator • Draft
6 of 16 tasks
feat: support client-supplied uuid for multimodal processor cache.
#2243 opened Aug 14, 2026 by wly-115 Collaborator Loading…
8 of 17 tasks
bugfix: fix tilelang g_tilingKey multiple definition across kernel families.
#2242 opened Aug 14, 2026 by fengz72 Contributor Loading…
8 of 17 tasks
feat: add DeepSeek V4 NPU kernel Python bindings.
#2241 opened Aug 14, 2026 by 0moyi0-2024 Loading…
23
feat: add opt-in FIA decode for Qwen3.5.
#2240 opened Aug 14, 2026 by Enguikong Contributor Loading…
12
feat: support DeepSeek-V4 in the Python model executor
#2238 opened Aug 14, 2026 by 0moyi0-2024 Loading…
refactor: adopt SequenceParallelMixin for Wan2.2 sequence parallelism and communication-coputation overlap.
#2237 opened Aug 14, 2026 by zhangjieshao98 Contributor Loading…
feat: add sequence parallel and cache support for Flux DiT.
#2236 opened Aug 13, 2026 by z-jun03 Collaborator Loading…
8 of 17 tasks
feat: add MUSA CICD supports.
#2235 opened Aug 13, 2026 by fay85 Collaborator Loading…
4
fix: avoid unnecessary brpc recompilation.
#2232 opened Aug 13, 2026 by Kang-Meng Collaborator Loading…
17 tasks
refactor: sp refactor for flux2 dit module on device npu
#2229 opened Aug 13, 2026 by wang-shuibin Collaborator Loading…
9 of 17 tasks
feat: bridge DeepSeek V4 Python caches and metadata.
#2228 opened Aug 13, 2026 by 0moyi0-2024 Loading…
1
feat: enable dit cache for wan dual transformers.
#2226 opened Aug 13, 2026 by z-jun03 Collaborator Loading…
3 of 8 tasks
feat: change the aclnnAdalayernorm to v2 && move the cast for aclnnlayernorm.
#2224 opened Aug 13, 2026 by ethan686 Collaborator Loading…
ProTip! Add no:assignee to see everything that’s not assigned.