Skip to content

Pull requests: ROCm/vllm

New pull request New
34 Open 821 Closed
34 Open 821 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

[kernels] tune NUM_SEGMENTS, attn_warps for TRITON_ATTN 3d kernel on gfx1151
#887 opened Apr 17, 2026 by amd-callumm Loading…
5 tasks done
Skinny int4 perf
#886 opened Apr 17, 2026 by eble-amd • Draft
3 of 5 tasks
[ROCm][Refactor] Enable GPTQMarlinConfig on ROCm to use choose_mp_linear_kernel
#885 opened Apr 17, 2026 by mgehre-amd Loading…
@eble-amd
Wrap triton_scaled_mm in an opaque custom op
#883 opened Apr 17, 2026 by mgehre-amd Loading…
gfx11: Merge upstream/main (2026-04-17)
#882 opened Apr 17, 2026 by mgehre-amd • Draft
Tune hybrid_triton_w4a16 prefill kernel for gfx1151
#879 opened Apr 15, 2026 by mgehre-amd • Draft
3 tasks done
Use fp16 fdot2 for bf16 int4 GEMV dequant on RDNA 3.5
#877 opened Apr 15, 2026 by mgehre-amd • Draft
Add hybrid MoE kernel with wvSplitK int4 GEMM
#876 opened Apr 15, 2026 by mgehre-amd Loading…
5 tasks done
Fix Triton WNA16 MoE fallback for CompressedTensorsWNA16MoEMethod
#875 opened Apr 15, 2026 by mgehre-amd • Draft
2
Upload vllm wheels to PEP 503 simple index on S3
#874 opened Apr 14, 2026 by mgehre-amd Loading…
1
Add gfx12 (RDNA4) Triton tile-size heuristic for W4A16 prefill kernel
#870 opened Apr 13, 2026 by amd-xavierwang Loading…
3 of 5 tasks
2
Enable FLASH_ATTN backend with upstream flash-attn CK on ROCm
#866 opened Apr 10, 2026 by mgehre-amd • Draft
1 task
update rocm version truncate
#864 opened Mar 16, 2026 by kiran-thumma Loading…
1
[CI/Build] Fix Dockerfile.rocm_base image build for ROCm 7.2 bug Something isn't working
#863 opened Feb 23, 2026 by jbelloncastro Loading…
3 of 5 tasks
1
Create devcontainer.json
#859 opened Feb 11, 2026 by vagabond2522 Loading…
5 tasks
adding cli entry for moe_configs
#858 opened Feb 11, 2026 by danichan-mkm Loading…
1
Add silu-and-mul and per-token dynamic FP8 quant fusion stale
#852 opened Jan 6, 2026 by kliuae-amd Loading…
5 tasks
1
[Triton] fix rope kv_cache for RocmAiterUnifiedAttentionImpl + v.contiguous() remove stale
#851 opened Jan 2, 2026 by k50112113 Loading…
1
[feat] Add ROCm ATOM model impl backend
#850 opened Dec 25, 2025 by zejunchen-zejun • Draft
Update unquantized_fused_moe_method.py stale
#849 opened Dec 24, 2025 by aaab8b Loading…
5 tasks
2
optimized dsv3.2 missing underscore
#847 opened Dec 18, 2025 by gronsti-amd Loading…
5 tasks
2
Amd/deepseek aiter fusions
#842 opened Dec 8, 2025 by dllehr-amd Collaborator • Draft
5 tasks
[Triton] Add Fused GEMM A8W8 + Split + Concat to Deepseek FP8 Model
#831 opened Dec 3, 2025 by farlukas • Draft
1 of 3 tasks
[rocm]use aiter triton kernel as triton mha fallback path
#809 opened Nov 14, 2025 by zhuyuhua-v • Draft
1
add aiter fusion pattern for sequence parallel
#781 opened Oct 31, 2025 by zhuyuhua-v • Draft
5 tasks
ProTip! Filter pull requests by the default branch with base:main.