Skip to content

Pull requests: ROCm/vllm

New pull request New
40 Open 980 Closed
40 Open 980 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

Release/0.23.0
#1058 opened Jul 14, 2026 by kiran-thumma Collaborator • Draft
5 tasks
Use AITER Backend for Dsv4
#1057 opened Jul 13, 2026 by JadenMathias Loading…
Patch for mori build
#1052 opened Jul 9, 2026 by JadenMathias Loading…
[ROCm][Kernel] W4A16 skinny GEMM: hand-asm kernel for gfx1151
#1045 opened Jul 1, 2026 by mgehre-amd • Draft
[bench] vit_attention/bench.py: add --backend, default to running all
#1044 opened Jul 1, 2026 by mgehre-amd • Draft
[CI] Re-enable performance test job
#1033 opened Jun 29, 2026 by marcusr-amd • Draft
5 tasks
1
[ROCm] Split skinny_gemms_int8.cu into per-N translation units
#1028 opened Jun 26, 2026 by marcusr-amd • Draft
3 of 5 tasks
1
tune Qwen3-VL-4B prefill unified-attention on gfx1150
#1024 opened Jun 26, 2026 by qingxuamd Loading…
Fix GLM 5.2 mxfp4 MTP loading issue
#1022 opened Jun 25, 2026 by amd-xiaoyu12 Loading…
[ROCm][MoE] W4A16 MoE routing-distribution benchmark suite for the gfx11 prefill GEMM
#1020 opened Jun 25, 2026 by roberteg16 • Draft
optimize TTFT qwen3-vl
#1006 opened Jun 15, 2026 by qingxuamd Loading…
1
455 war room findings
#1001 opened Jun 12, 2026 by jpvillam-amd Loading…
@jpvillam-amd
MoE: Grouped Triton GEMM for TTFT improvements
#970 opened May 26, 2026 by mgehre-amd • Draft
[ROCm][MoE] Modular MoE: alias fused_out with output to skip finalize copy
#940 opened May 19, 2026 by mgehre-amd Loading…
2 tasks done
10
feat: Add NPU+GPU async pipelining for vision-language models
#936 opened May 14, 2026 by liangliangchang • Draft
4 of 5 tasks
10
Annotate VLM/audio tower nn.Linear calls in PyTorch profiles
#934 opened May 13, 2026 by mgehre-amd Loading…
Hybrid
#918 opened May 4, 2026 by liangliangchang • Draft
5 tasks
1
Auto-build flash-attn wheels on push, upload to S3
#910 opened Apr 30, 2026 by mgehre-amd • Draft
1 task
[ROCm][DSv4] Share AITER decode dequant + fp8-cast buffers across layers (rebased, stacked on #902)
#903 opened Apr 27, 2026 by ChuanLi1101 • Draft
2 of 4 tasks
4
[ROCm][DSv4] Make AITER sparse decode cudagraph-clean (rebased, stacked on #901)
#902 opened Apr 27, 2026 by ChuanLi1101 • Draft
2 of 5 tasks
5
[ROCm][DSv4] AITER-accelerated MLA decode for DeepSeek V4 on MI355X (rebased on tj/dsv4prrebase)
#901 opened Apr 27, 2026 by ChuanLi1101 • Draft
1 of 4 tasks
5
[Do Not Merge] For review purpose: Rocm/aiter mla dsv4 decode cudagraph
#900 opened Apr 26, 2026 by tjtanaavllm • Draft
5 tasks
[ROCm] support topk_softplus for all number of experts
#899 opened Apr 25, 2026 by tjtanaa Loading…
5 tasks
Tune hybrid_triton_w4a16 prefill kernel for gfx1151
#879 opened Apr 15, 2026 by mgehre-amd • Draft
3 tasks done
Enable FLASH_ATTN backend with upstream flash-attn CK on ROCm for decode
#866 opened Apr 10, 2026 by mgehre-amd • Draft
1 task
ProTip! Updated in the last three days: updated:>2026-07-10.