Skip to content

Pull requests: sgl-project/sglang

New pull request New
305 Open 3,628 Closed
305 Open 3,628 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
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

Remove unecessary is_fa3_supported check
#6112 opened May 8, 2025 by hebiao064 • Draft
6 tasks
[Feat] Support FlashMLA backend with MTP and FP8 KV cache
#6109 opened May 8, 2025 by quinnrong94 Loading…
6 tasks
@Fridge003
Tiny refactor bench_serving
#6108 opened May 8, 2025 by fzyzcjy Loading…
6 tasks
Support outputing details for bench_serving
#6107 opened May 8, 2025 by fzyzcjy Loading…
6 tasks
Support VILA models
#6106 opened May 8, 2025 by futrime • Draft
6 tasks
Support precomputed multimodal features for qwen-vl models. high priority visIon-LM
#6103 opened May 7, 2025 by ysulsky • Draft
1 of 6 tasks
1
Cutlass MLA: Disable split kv due to https://github.com/NVIDIA/cutlass/issues/2274
#6101 opened May 7, 2025 by trevor-m Loading…
6 tasks
1
[Metrics] Add KV events publishing
#6098 opened May 7, 2025 by trevor-m Loading…
6 tasks
2
[AMD] switch to custom allreduce regardless of MSCCL setting on ROCm
#6097 opened May 7, 2025 by hubertlu-tw Loading…
6 tasks
Log the timestamps of each prefill/decode iteration
#6094 opened May 7, 2025 by yuhsuan-t Loading…
1 of 6 tasks
[1/2] Add Kernel support for Cutlass based Fused FP4 MoE
#6093 opened May 7, 2025 by pavanimajety • Draft
6 tasks
[Performance] Use the max num_tokens per DP rank as the CUDA graph batch size
#6092 opened May 7, 2025 by xpmemeda Loading…
1 of 6 tasks
@zhyncs
2
Change AMD test threshold
#6091 opened May 7, 2025 by fzyzcjy Loading…
6 tasks
sampling_params: add thinking_budget
#6089 opened May 7, 2025 by thyecust Loading…
1 of 6 tasks
@minleminzui
1
feat: mtp support dp-attention
#6081 opened May 7, 2025 by u4lr451 Loading…
1 of 6 tasks
1
@fzyzcjy @merrymercy
@ispobock @ch-wan
1
chore: upgrade deepgemm
#6073 opened May 7, 2025 by zhyncs Loading…
6 tasks
@Alcanderian
2
Refactor DeepSeek model for two-batch overlap
#6070 opened May 7, 2025 by fzyzcjy Loading…
6 tasks
[Feat] Integrate flashmla into sgl-kernel
#6065 opened May 7, 2025 by PopSoda2002 Loading…
1 task done
7
[Intel][Gaudi] Remove HPU's dependency on vllm
#6061 opened May 6, 2025 by yangw1234 • Draft
6 tasks
Support XiaomiMiMo inference with mtp
#6059 opened May 6, 2025 by ryang-max Loading…
1 of 6 tasks
2
Feat/support rerank
#6058 opened May 6, 2025 by woodx9 Loading…
2 of 6 tasks
3
Fix is_all_greedy flag update when merging batches high priority
#6055 opened May 6, 2025 by YavorGIvanov Loading…
6 tasks done
fix typo in spec decoding
#6053 opened May 6, 2025 by RuixiangMa Loading…
restruct fused moe code style to avoid long lines files
#6052 opened May 6, 2025 by BBuf Loading…
6 tasks
[Docs] Minor changes to development_guide_using_docker.md
#6049 opened May 6, 2025 by windsonsea Loading…
3 of 6 tasks
ProTip! Add no:assignee to see everything that’s not assigned.