Skip to content

Pull requests: vllm-project/vllm-gaudi

New pull request New
39 Open 1,372 Closed
39 Open 1,372 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

Draft for #1537
#1545 opened Jun 12, 2026 by iboiko-habana Collaborator Loading…
6
Update Dockerfiles and documentation for v0.19.1.post1 release documentation Improvements or additions to documentation skip-gaudi-tests
#1544 opened Jun 12, 2026 by PatrykWo Collaborator Loading…
2
Reapply path for block_size for non-GDN hybrid models
#1543 opened Jun 11, 2026 by rsmyrek Contributor Loading…
4
Fix contiguous PA KV fetch reading wrong physical rows at long context
#1540 opened Jun 10, 2026 by adobrzyn Collaborator • Draft
2
Add Qwen3.6-35B CI test
#1537 opened Jun 9, 2026 by yeonsily Contributor Loading…
9
[FIX_FOR_VLLM_CUSTOM=d841386d272200dd381a5791833771db9a47adf7] Adapt HPU MoE to the FusedMoE/MoERunner inversion refactor
#1536 opened Jun 9, 2026 by pawel-olejniczak Collaborator Loading…
3
Reapply path for block_size for non-GDN hybrid models
#1535 opened Jun 9, 2026 by rsmyrek Contributor Loading…
3
Added Hi Cache CPU Offloading test to CI
#1533 opened Jun 8, 2026 by slokesha Contributor • Draft
5
Revert "Port of: fix: hybrid model warmup block_size mismatch (Qwen3.…
#1532 opened Jun 8, 2026 by rsmyrek Contributor • Draft
4
[WIP][GDN] Integrate gdn_read_decayed_state TPC kernel into decode path
#1531 opened Jun 8, 2026 by osavchenkox Contributor Loading…
4 of 6 tasks
11
Add Single Card support in GPTOSS
#1526 opened Jun 4, 2026 by SKRohit Contributor • Draft
9
Reapply path for block_size for non-GDN hybrid models
#1523 opened Jun 3, 2026 by ksmusz Contributor Loading…
3
dynamic_quant: 20% improvements in per-channel mode
#1512 opened Jun 1, 2026 by osavchenkox Contributor Loading…
6
fix: raise default max_cudagraph_capture_size floor to 16384
#1502 opened May 27, 2026 by kamil-kaczor Contributor Loading…
1 of 3 tasks
5
fix: prevent eager-mode env vars leaking to lazy-mode subprocesses
#1501 opened May 27, 2026 by kamil-kaczor Contributor Loading…
3 of 4 tasks
5
[DO_NOT_MERGE] Skip gather/index_copy for sequential state indices in GDN decode
#1487 opened May 24, 2026 by osavchenkox Contributor Loading…
4
[DO_NOT_MERGE] Use HPU causal_conv1d_update TPC kernel instead of multiple ops
#1484 opened May 23, 2026 by osavchenkox Contributor Loading…
14
[DO_NOT_MERGE] Remove repeat_interleave from GDN recurrent path
#1483 opened May 22, 2026 by osavchenkox Contributor Loading…
5
fix(test): work around Gaudi3 sdpa_recomp_fwd GQA kernel bug in reference
#1467 opened May 20, 2026 by yangulei Collaborator • Draft
5
[DO_NOT_MERGE] Use l2Norm TPC kernel for HPU compile
#1457 opened May 19, 2026 by osavchenkox Contributor Loading…
9
fix unit_moe: use loaded weight scales, force activation scale=1.0
#1456 opened May 19, 2026 by yangulei Collaborator • Draft
4
Global patch for t.compile high warmup time on MoE models
#1452 opened May 17, 2026 by tvoas Contributor Loading…
19
[WIP] enable step3p5 for bf16 and fp8
#1382 opened Apr 22, 2026 by ranzhejiang • Draft
36
Enhance GDN compact mode with prefix caching support
#1369 opened Apr 17, 2026 by jinyouzhi Contributor • Draft
6
Fix yapf formatting in test_bucketing_issue_1347.py from PR #1348
#1355 opened Apr 15, 2026 by Copilot AI • Draft
3 tasks done
@michalkuligowski
ProTip! Follow long discussions with comments:>50.