-
Notifications
You must be signed in to change notification settings - Fork 5.6k
Pull requests: sgl-project/sglang
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[fix][PP][NPU] Fix PP gibberish output due to proxy tensor TP all-gather mixing
#23816
opened Apr 27, 2026 by
litmei
Contributor
Loading…
3 of 5 tasks
[NPU] Fix DeepEP LL dispatch BF16 flag and skip triton kernel on NPU for Qwen3.5
#23815
opened Apr 27, 2026 by
iridiumine
Contributor
Loading…
3 of 5 tasks
Cherry pick clipping logic from https://github.com/sgl-project/sglang…
#23813
opened Apr 27, 2026 by
ovidiusm
Contributor
Loading…
[Feature] Xiaomi MiMo-V2-Omni day0 support
high priority
Multi-modal
multi-modal language model
new-model
run-ci
#23811
opened Apr 27, 2026 by
Abatom
Contributor
Loading…
5 tasks
fix act fun for xpu
diffusion
SGLang Diffusion
#23809
opened Apr 27, 2026 by
sushildubey171
Contributor
Loading…
5 tasks
[Feature] Xiaomi MiMo-V2-Pro day0 support
run-ci
#23808
opened Apr 27, 2026 by
JoyFuture
Contributor
Loading…
5 tasks
Fix Request.is_disconnected() broken by metrics middleware (issue #15686)
#23807
opened Apr 27, 2026 by
brucechanglongxu
Contributor
Loading…
fix: guard against None new_accepted_tokens in vocab boundary check
#23806
opened Apr 27, 2026 by
Ricardo-M-L
Contributor
Loading…
fix: normalize role key to lowercase in janus-pro chat template
#23805
opened Apr 27, 2026 by
Ricardo-M-L
Contributor
Loading…
2 tasks
[tokenizer] Guard communicator handle_recv against stale messages
#23804
opened Apr 27, 2026 by
brucechanglongxu
Contributor
Loading…
fix: stop-string check misses early matches during speculative decoding
speculative-decoding
#23802
opened Apr 27, 2026 by
xythink
Loading…
2 tasks done
docs: update Python prerequisite to 3.10
documentation
Improvements or additions to documentation
#23801
opened Apr 27, 2026 by
Kare0638
Contributor
Loading…
5 tasks done
[DeepSeek V4] Use get_rope_config for MQALayer RoPE initialization
deepseek
#23800
opened Apr 27, 2026 by
liquanfeng
Contributor
Loading…
2 tasks
docs(Qwen3.6): add serving benchmark results for 35B-A3B-FP8
#23798
opened Apr 27, 2026 by
prakashkagitha
•
Draft
4 of 5 tasks
Skip LM head during FlashInfer autotune dummy run
#23796
opened Apr 27, 2026 by
Kangyan-Zhou
Collaborator
•
Draft
4 tasks
🚧 [llm][npu][quant] Add W4A4 MXFP4 quantization support for Qwen3 Dense on Ascend NPU
diffusion
SGLang Diffusion
npu
quant
LLM Quantization
#23795
opened Apr 27, 2026 by
TallMessiWu
Loading…
[DO NOT MERGE] DeepSeek V4 Dev Branch
#23791
opened Apr 27, 2026 by
Fridge003
Collaborator
Loading…
5 tasks
[BugFix] Fix AttributeError in eagle speculative decoding when bs != len(sampling_info)
#23790
opened Apr 27, 2026 by
lingebeng
Contributor
Loading…
fix(deepseekv32): parse bare invoke blocks without function_calls wrapper
deepseek
#23786
opened Apr 27, 2026 by
Kangyan-Zhou
Collaborator
•
Draft
4 of 5 tasks
Add WLFU radix cache eviction policy
#23781
opened Apr 27, 2026 by
tejas-goyal
Loading…
4 of 5 tasks
[Perf] Optimize Triton RMSNorm to fix register spilling and low occupancy on mid-tier GPUs
jit-kernel
#23775
opened Apr 26, 2026 by
Umang-projects
Loading…
Previous Next
ProTip!
Filter pull requests by the default branch with base:main.