Skip to content

Pull requests: jd-opensource/xllm

New pull request New
52 Open 949 Closed
52 Open 949 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

perf: optimize qwen3.5 hybrid linear cache flow[4/N].
#1160 opened Apr 1, 2026 by JC-ut0 Loading…
4
refactor: introduce MultimodalProcessor abstraction for VLM.
#1159 opened Apr 1, 2026 by wly-115 • Draft
10
feat: add constrain decodeing for onerec.
#1158 opened Apr 1, 2026 by DragonFive Loading…
16
refactor: prefer lambda to bind.
#1157 opened Apr 1, 2026 by Dragonliu2018 Loading…
12
feat: support video inference for Qwen3-VL on NPU device.
#1151 opened Mar 31, 2026 by xanecdotex Loading…
10
bugfix: make CP compatible with MTP
#1150 opened Mar 31, 2026 by shifengmin Loading…
14
feat: support FIA for qwen model on npu device.
#1147 opened Mar 31, 2026 by sanlio36 Loading…
8
bugfix: rollback shared prefix blocks on allocate failure.
#1146 opened Mar 31, 2026 by RobbieLeung Loading…
3
feat: implement column parallel for lm head to improve performance.
#1145 opened Mar 31, 2026 by wxh571001500 Loading…
13
feat: support flux model on mlu device.
#1138 opened Mar 30, 2026 by phantomlei3 Loading…
2
feat: support embedding interface for all generate VLM models.
#1136 opened Mar 30, 2026 by xanecdotex Loading…
3
feat: support oxygenvlm model on mlu device.
#1131 opened Mar 30, 2026 by phantomlei3 Loading…
15
perf: optimize qwen3.5 hybrid linear cache flow[4/N].
#1130 opened Mar 30, 2026 by yingxudeng Loading…
4
feat: support configurable max/min pixels for vlm image processors.
#1123 opened Mar 27, 2026 by wly-115 Loading…
3
feat: support joyai-llm-flash model on npu device.
#1121 opened Mar 27, 2026 by longhui-z Loading…
10
feat: support parsing dtype field from config.
#1118 opened Mar 26, 2026 by DongheJin Loading…
2
refactor: simplify xllm_ops pre-build checks and marker-based rebuild logic.
#1111 opened Mar 25, 2026 by LMX-xin Loading…
7
feat: support pured lm head by candidate token ids.
#1071 opened Mar 17, 2026 by RobbieLeung Loading…
feat: support qwen3-omni talker and code2wav.
#1070 opened Mar 17, 2026 by ethan686 Loading…
10
[WIP] feat: add functional MiniMax-M2.5 baseline
#1064 opened Mar 16, 2026 by QwertyJack • Draft
4
feat: add onerec in supported model docs and align rec utility style.
#1055 opened Mar 13, 2026 by DragonFive Loading…
4
feat: add onerec model implement[4/N].
#1051 opened Mar 13, 2026 by DragonFive Loading…
3 of 6 tasks
7
feat: add onerec model implement[3/N].
#1050 opened Mar 13, 2026 by DragonFive Loading…
5 of 9 tasks
5
bugfix: avoid decode instance leak if prefill instance prefill fail.
#1046 opened Mar 12, 2026 by magicheng0816 Loading…
4
bugfix: fix decode instance infinite retry allocate.
#1045 opened Mar 12, 2026 by magicheng0816 Loading…
4
ProTip! Follow long discussions with comments:>50.