Skip to content

Pull requests: vllm-project/vllm-gaudi

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
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort

Pull requests list

Optimize performance of static FP8 linear op
#715 opened Dec 13, 2025 by skavulya Loading…
Add heterogeneous pd docs
#714 opened Dec 13, 2025 by pi314ever Draft
Create UBI based vLLM docker build instructions documentation Improvements or additions to documentation skip-gaudi-tests
#713 opened Dec 12, 2025 by ghandoura Loading…
Add ucx test
#711 opened Dec 12, 2025 by pi314ever Draft
Fix for Llama4 static quantization
#707 opened Dec 10, 2025 by vidyasiv Loading…
Fix missing unique attn image in the docs
#706 opened Dec 9, 2025 by leopck Loading…
Unified attn FP8 perf optimizations
#705 opened Dec 9, 2025 by afierka-intel Loading…
qwen3-vl enablement
#700 opened Dec 8, 2025 by libinta Loading…
Qwen2.5 vl no alignment
#698 opened Dec 8, 2025 by slokesha Loading…
Add 0.12.0 release notes documentation Improvements or additions to documentation skip-gaudi-tests
#694 opened Dec 5, 2025 by mhelf-intel Loading…
Fix the docker image path documentation Improvements or additions to documentation skip-gaudi-tests
#691 opened Dec 5, 2025 by mhelf-intel Loading…
Add support for chunked attention (#597)
#683 opened Dec 4, 2025 by jkaniecki Loading…
Add support for chunked attention (#597)
#682 opened Dec 4, 2025 by jkaniecki Loading…
Initiate CI with libfabric backend
#679 opened Dec 4, 2025 by amathewc Loading…
Add local path option for hf_cache
#662 opened Dec 1, 2025 by PatrykWo Loading…
CustomOp: grouped topk
#647 opened Nov 27, 2025 by xinyu-intel Loading…
bucket: add query len 1 to prefill bucket
#645 opened Nov 27, 2025 by xinyu-intel Loading…
Hybrid KV cache for hpu
#644 opened Nov 26, 2025 by michalkuligowski Draft
ProTip! Find all pull requests that aren't related to any open issues with -linked:issue.