-
Notifications
You must be signed in to change notification settings - Fork 240
Pull requests: jd-opensource/xllm
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
test: support npu a3 github ci workflow.
#1849
opened Jun 29, 2026 by
ksk0014
Loading…
9 of 17 tasks
feat: enable SwiGluQuant fusion for Qwen3 MLP
#1847
opened Jun 29, 2026 by
nie-linfeng
Contributor
Loading…
feat: add linear-state block type and sequence slot accessors (3/n).
#1846
opened Jun 29, 2026 by
yingxudeng
Collaborator
Loading…
17 tasks
bugfix: propagate is_hybrid_linear_attention for Qwen3.5 VLM graph capture
#1844
opened Jun 29, 2026 by
maojunx99
Contributor
Loading…
3 of 11 tasks
bugfix: fix num_used_block collections in pd prefill.
#1841
opened Jun 29, 2026 by
phantomlei3
Collaborator
Loading…
8 of 17 tasks
[WIP]feat: support qwen3.5 linear-state prefix cache.
#1839
opened Jun 27, 2026 by
yingxudeng
Collaborator
•
Draft
17 tasks
bugfix: synchronize MTP compute stream after draft/validate forward.
#1838
opened Jun 27, 2026 by
yingxudeng
Collaborator
Loading…
17 tasks
feat: add beam search torch support to mlu.
#1835
opened Jun 27, 2026 by
phantomlei3
Collaborator
•
Draft
17 tasks
feat: support batch embedding requests.
#1828
opened Jun 26, 2026 by
DongheJin
Collaborator
Loading…
17 tasks
bugfix: fix MTP DP concurrent inference failures.
#1817
opened Jun 24, 2026 by
DongheJin
Collaborator
Loading…
17 tasks
feat: enable bf16 fallback for o_proj in Qwen3-VL W8A8 quantization
#1809
opened Jun 23, 2026 by
nie-linfeng
Contributor
Loading…
feat: add scheduler-side linear-state prefix cache and sequence slot (3/n).
#1806
opened Jun 23, 2026 by
yingxudeng
Collaborator
Loading…
17 tasks
bugfix: fix DeepSeek V4 schedule_overlap + mtp.
#1782
opened Jun 18, 2026 by
JC-ut0
Contributor
Loading…
17 tasks
bugfix: resolve cross-NUMA spawn worker isolation issues
#1776
opened Jun 18, 2026 by
asr-sheep1
Collaborator
Loading…
8 of 17 tasks
feat: support eagle3 for vlm models.
#1773
opened Jun 17, 2026 by
shan-chen-feng
Collaborator
Loading…
17 tasks
Conv1d decode for release0.10.0
#1767
opened Jun 17, 2026 by
BikingNow
Contributor
Loading…
4 of 17 tasks
Switch TileLang Ascend kernels to PTO backend
#1765
opened Jun 17, 2026 by
ShareableXue
Loading…
3 of 7 tasks
[WIP] feat: support qwen3.5 linear-state prefix cache (remaining).
#1755
opened Jun 16, 2026 by
yingxudeng
Collaborator
•
Draft
17 tasks
feat: support MiniMax-M3 on npu device.
#1731
opened Jun 13, 2026 by
QwertyJack
Collaborator
Loading…
bugfix: fix dtype mismatch between scatter src and dst.
#1701
opened Jun 11, 2026 by
shifengmin
Collaborator
Loading…
7 of 17 tasks
bugfix: fix graph decode warmup with MTP.
#1694
opened Jun 11, 2026 by
shifengmin
Collaborator
Loading…
7 of 17 tasks
Previous Next
ProTip!
Type g i on any issue or pull request to go back to the issue listing page.