Upstream sync 22/N: merge 81e81dac1f (75 commits, conflict-free) by roberteg16 · Pull Request #1269 · ROCm/vllm

roberteg16 · 2026-09-07T09:18:52Z

Summary

Conflict-free upstream batch: the 75 commits between 50ba4bc6b2 and 81e81dac1f (2026-08-13 15:44 UTC .. 2026-08-14 22:11 UTC). git merge reported no conflicts.

391 files, +17875/−3220. No deleted files.

This batch is a deliberate split. The ladder found a clean run of 151
commits, over the ~150 threshold: a single PR that large has no useful bisection point if it regresses, and at 638 files it is not reviewable. Split at k=75; batch 23 carries the remaining 76 commits of the same clean run.

Boundary probed in a ladder: k=25/100/150 clean, k=200 conflicts, k=300/400/ 500 and the tip (k=599) all conflict; stepping forward, k=151 is the last clean commit and k=152 conflicts.

Merge commit only — do not squash or rebase.

AI assistance was used to prepare this merge.

Test plan

pre-commit run --all-files green on the merged tree
Build + 6-model benchmark sweep vs the Strix Halo dashboard — no regression on transformers 5.15.0, all six within the gate (results)

…-project#51862) Signed-off-by: kliuae <kuanfu.liu@embeddedllm.com>

…acts (vllm-project#52127) Signed-off-by: jiang1.li <jiang1.li@intel.com>

…llm-project#52134) Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

…llm-project#49458) Signed-off-by: Thomas Ortner <boh@zurich.ibm.com>

…llm-project#50221) Signed-off-by: jperezde <jperezde@redhat.com>

…vllm-project#51611) Signed-off-by: qwerqwerqwe8688-jpg <xuuuuu2021@163.com>

Signed-off-by: Jason Ozuzu <jasonozuzu@cohere.com>

…2021) Signed-off-by: Taneem Ibrahim <taneem.ibrahim@gmail.com>

…#50534) Signed-off-by: pmanczak <pawel.manczak@intel.com>

…t#52145) Signed-off-by: Vineeta Tiwari <vineeta.tiwari2@ibm.com> Co-authored-by: Vineeta Tiwari <vineeta.tiwari2@ibm.com>

Signed-off-by: ruirui6946 <142162413+ruirui6946@users.noreply.github.com>

…llm-project#52114) Signed-off-by: zexplorerhj <zhjoneson@163.com>

Signed-off-by: bk-201 <joy25810@foxmail.com> Signed-off-by: linitra24 <Joy25810@foxmail.com> Signed-off-by: linitra24 <joy25810@gmail.com> Co-authored-by: Jee Jee Li <pandaleefree@gmail.com> Co-authored-by: linitra24 <joy25810@gmail.com>

) Signed-off-by: KurodaKanbei <mistergalahad@gmail.com>

…m-project#52030) Signed-off-by: mgoin <mgoin64@gmail.com>

Signed-off-by: vllmellm <vllm.ellm@embeddedllm.com>

Signed-off-by: aoshen02 <aoshen@inferact.ai> Co-authored-by: OpenAI Codex <codex@openai.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>

…t#52147) Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

… `create_engine_config` (vllm-project#47692) Signed-off-by: Syed Ali Jaseem <syedali0057@gmail.com> Co-authored-by: Claude <noreply@anthropic.com>

…ing VideoDummyOptions to _get_dummy_images (vllm-project#52122) Signed-off-by: mayuyuace <qiming1.zhang@intel.com>

Signed-off-by: Thien Tran <gau.nernst@yahoo.com.sg>

…ct#52173) Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

…t#52171) Signed-off-by: Nick Iusiumbeli <nickuspro@gmail.com>

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>

Signed-off-by: Andreas Karatzas <Andreas.Karatzas@amd.com> Signed-off-by: Andreas Karatzas <akaratza@amd.com>

…project#51793) Signed-off-by: Felix Marty <Felix.Marty@amd.com>

…l grammar terminates (vllm-project#50595) Signed-off-by: Yifan Zong <yzong@redhat.com>

…_utils import tilelang` and relaxed `has_tilelang` (vllm-project#51159) Signed-off-by: Felix Marty <Felix.Marty@amd.com> Co-authored-by: Andreas Karatzas <akaratza@amd.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>

…ms (vllm-project#52139) Signed-off-by: Stefan Koncarevic <Stefan.Koncarevic@amd.com>

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>

…omize_spec (vllm-project#51704) Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com> Co-authored-by: Claude <noreply@anthropic.com>

Signed-off-by: mayuyuace <qiming1.zhang@intel.com> Signed-off-by: Andreas Karatzas <Andreas.Karatzas@amd.com> Co-authored-by: Andreas Karatzas <Andreas.Karatzas@amd.com>

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>

…DFlash aux state (vllm-project#50487) Signed-off-by: Rahul Chalamala <22563365+rchalamala@users.noreply.github.com> Co-authored-by: Janelle Cai <janelle.cai@modal.com>

…t abort engine startup (vllm-project#52241) Signed-off-by: shanjiaz <zsjwpianpian@gmail.com>

Signed-off-by: Stefan Koncarevic <stefan.koncarevic@amd.com>

…llm-project#51216) Signed-off-by: jamesETsmith <james.smith9113@gmail.com> Co-authored-by: Andreas Karatzas <akaratza@amd.com>

…t#52369) Signed-off-by: Nick Hill <nickhill123@gmail.com>

Signed-off-by: Connor Carpenter <connorc@nvidia.com> Co-authored-by: Nick Hill <nickhill123@gmail.com>

…oject#50597) Signed-off-by: Stacy Roberts <sroberts@amd.com> Signed-off-by: Stacy Roberts <robers23@utexas.edu> Co-authored-by: Stacy Roberts <robers23@utexas.edu>

…project#50589) Signed-off-by: Guanxin Li <guanxinl@nvidia.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

…llm-project#52328)

Signed-off-by: Kevin Luu <51931015+khluu@users.noreply.github.com> Signed-off-by: khluu <khluu000@gmail.com>

Signed-off-by: Kevin Luu <51931015+khluu@users.noreply.github.com>

kliuae and others added 30 commits August 13, 2026 09:11

[ROCm][Perf] Kimi-K3 Remove prefill pipeline stall in chunk KDA (vllm…

f962616

…-project#51862) Signed-off-by: kliuae <kuanfu.liu@embeddedllm.com>

[CI/Build][CPU] Shrink triton-cpu-build layer by dropping build artif…

1c3633a

…acts (vllm-project#52127) Signed-off-by: jiang1.li <jiang1.li@intel.com>

[Docs] Fix WhisperEncoderLayer.forward docstring in dots3_note (v…

903da60

…llm-project#52134) Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

Hardware-agnostic model definition via HF transformer backend (1/N) (v…

b8baa31

…llm-project#49458) Signed-off-by: Thomas Ortner <boh@zurich.ibm.com>

fix(security): enforce audio decode duration limit in NanoNemotronVL (v…

37c3bdf

…llm-project#50221) Signed-off-by: jperezde <jperezde@redhat.com>

[Doc] Fix stale rejection_sample_method and synthetic_acceptance_rate (…

443fa5a

…vllm-project#51611) Signed-off-by: qwerqwerqwe8688-jpg <xuuuuu2021@163.com>

chore: Upstream Cohere parser fixes + tests (vllm-project#51998)

5fee0a8

Signed-off-by: Jason Ozuzu <jasonozuzu@cohere.com>

[Bugfix] Preserve Anthropic disable_parallel_tool_use (vllm-project#5…

d0ae25e

…2021) Signed-off-by: Taneem Ibrahim <taneem.ibrahim@gmail.com>

[XPU] Add tuned Mamba SSU configs for Intel Arc Pro B70 (vllm-project…

c4e9692

…#50534) Signed-off-by: pmanczak <pawel.manczak@intel.com>

[Misc] Add missing return type annotations in outputs.py (vllm-projec…

015660d

…t#52145) Signed-off-by: Vineeta Tiwari <vineeta.tiwari2@ibm.com> Co-authored-by: Vineeta Tiwari <vineeta.tiwari2@ibm.com>

[Frontend] Log output token IDs at DEBUG level (vllm-project#52098)

152c913

Signed-off-by: ruirui6946 <142162413+ruirui6946@users.noreply.github.com>

[Model] [Quantization] Add Ling hybrid MXFP4 routed experts support (v…

8e1131e

…llm-project#52114) Signed-off-by: zexplorerhj <zhjoneson@163.com>

[Bugfix] Disable sequence parallelism for Dots3 NOTE (vllm-project#52172

170592a

) Signed-off-by: KurodaKanbei <mistergalahad@gmail.com>

[Bugfix] Fix packed GDN decode launch for large batch-head grids (vll…

2d24355

…m-project#52030) Signed-off-by: mgoin <mgoin64@gmail.com>

[ROCm] Enable V2 model runner for Kimi-K3 on ROCm (vllm-project#51653)

f3c1638

Signed-off-by: vllmellm <vllm.ellm@embeddedllm.com>

[Frontend] Add routed-experts prompt offset (vllm-project#51906)

7553aac

Signed-off-by: aoshen02 <aoshen@inferact.ai> Co-authored-by: OpenAI Codex <codex@openai.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>

Standardise weight tying on ParallelLMHead.tie_weights (vllm-projec…

80d6d55

…t#52147) Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

[Bugfix] Fix --data-parallel-start-rank 0 being treated as unset in…

64ca614

… `create_engine_config` (vllm-project#47692) Signed-off-by: Syed Ali Jaseem <syedali0057@gmail.com> Co-authored-by: Claude <noreply@anthropic.com>

[Bugfix][MiniCPM-V] Fix AssertionError in get_dummy_mm_data when pass…

96acd47

…ing VideoDummyOptions to _get_dummy_images (vllm-project#52122) Signed-off-by: mayuyuace <qiming1.zhang@intel.com>

[Kimi-K3] Add GEMM-RS for sequence parallelism (vllm-project#52079)

6014f9e

Signed-off-by: Thien Tran <gau.nernst@yahoo.com.sg>

Apply logit softcapping in Transformers modelling backend (vllm-proje…

b245d8e

…ct#52173) Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

[Bugfix] Declare SupportsEagle3 on KimiLinearForCausalLM (vllm-projec…

83d4c61

…t#52171) Signed-off-by: Nick Iusiumbeli <nickuspro@gmail.com>

Auto-ping Cohere on related issues (vllm-project#52091)

69d4c3a

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>

[ROCm][CI] Solidify entrypoint LLM lifecycle (vllm-project#51280)

b96bcd0

Signed-off-by: Andreas Karatzas <Andreas.Karatzas@amd.com> Signed-off-by: Andreas Karatzas <akaratza@amd.com>

[Quantization] Remove dead QuantizationConfig.is_mxfp4_quant (vllm-…

48825ac

…project#51793) Signed-off-by: Felix Marty <Felix.Marty@amd.com>

[Bugfix][Structured Output] Mask request stop tokens in xgrammar unti…

e6b2a8a

…l grammar terminates (vllm-project#50595) Signed-off-by: Yifan Zong <yzong@redhat.com>

[Bugfix][ROCm][CI] Give the AITER MLA decode metadata stub its MLA di…

11c3fa4

…ms (vllm-project#52139) Signed-off-by: Stefan Koncarevic <Stefan.Koncarevic@amd.com>

[Platform] Add check_runner_kv_caches_multi_layer (vllm-project#51633)

73b8394

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>

LucasWilkinson and others added 18 commits August 14, 2026 09:47

[5/N][KV-Cache Layout Refactor] Backend-published KV packing via cust…

57bd0ed

…omize_spec (vllm-project#51704) Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com> Co-authored-by: Claude <noreply@anthropic.com>

[UT][XPU] fix b12x UT (vllm-project#52265)

1f7427b

Signed-off-by: mayuyuace <qiming1.zhang@intel.com> Signed-off-by: Andreas Karatzas <Andreas.Karatzas@amd.com> Co-authored-by: Andreas Karatzas <Andreas.Karatzas@amd.com>

[Misc] Remove override_attention_dtype (vllm-project#48684)

cdc4824

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>

[Model][Spec Decode] Tap the pre-norm AttnRes mixture as the Kimi K3 …

03a8d0b

…DFlash aux state (vllm-project#50487) Signed-off-by: Rahul Chalamala <22563365+rchalamala@users.noreply.github.com> Co-authored-by: Janelle Cai <janelle.cai@modal.com>

[Bugfix] Widen flashinfer.comm import guard so a failed import doesn'…

e078a22

…t abort engine startup (vllm-project#52241) Signed-off-by: shanjiaz <zsjwpianpian@gmail.com>

[Test][LoRA] Speed up the LoRA test job (vllm-project#52331)

83ded8d

Signed-off-by: Stefan Koncarevic <stefan.koncarevic@amd.com>

[ROCm][AMD] Enable preshuffled sparse indexing for 16-token blocks (v…

9b0ab5d

…llm-project#51216) Signed-off-by: jamesETsmith <james.smith9113@gmail.com> Co-authored-by: Andreas Karatzas <akaratza@amd.com>

[Perf] Avoid more GPU<->CPU syncs in multimodal encoders (vllm-projec…

3e3ceb1

…t#52369) Signed-off-by: Nick Hill <nickhill123@gmail.com>

[Rust Frontend][gRPC] Add RL lifecycle control (vllm-project#51316)

f473870

Signed-off-by: Connor Carpenter <connorc@nvidia.com> Co-authored-by: Nick Hill <nickhill123@gmail.com>

[ROCm]Remove special-case SiTU support model-specific gating (vllm-pr…

c794754

…oject#50597) Signed-off-by: Stacy Roberts <sroberts@amd.com> Signed-off-by: Stacy Roberts <robers23@utexas.edu> Co-authored-by: Stacy Roberts <robers23@utexas.edu>

[CI][Test] Seed the DeepEP v2 MoE workers, not just the parent (vllm-…

925ea7e

…project#50589) Signed-off-by: Guanxin Li <guanxinl@nvidia.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

[CI] Shard Quantization job into 4 parallel shards (≤30 min target) (v…

d87ef45

…llm-project#52328)

[CI] Shard extended pooling model tests (vllm-project#52322)

549cef0

Signed-off-by: Kevin Luu <51931015+khluu@users.noreply.github.com> Signed-off-by: khluu <khluu000@gmail.com>

[CI] Shard multimodal extended generation 2 (vllm-project#52323)

694db07

Signed-off-by: Kevin Luu <51931015+khluu@users.noreply.github.com>

[CI] Shard MoE refactor B200 eval (vllm-project#52327)

81e81da

Signed-off-by: Kevin Luu <51931015+khluu@users.noreply.github.com>

Upstream sync 22/N: merge 81e81da (75 commits, conflict-free)

9bc0e6e

Merge batch branch -117 into -118 (absorb gfx11 c84454f)

b2a00fe

Merge batch branch -117 into -118 (skinny GEMM K-contiguity gate)

f320a33

roberteg16 mentioned this pull request Sep 8, 2026

Upstream sync 25/N: merge d785eb51ce (15 commits, conflict-free) #1272

Merged

2 tasks done

roberteg16 requested a review from mgehre-amd September 8, 2026 10:12

roberteg16 marked this pull request as ready for review September 8, 2026 10:12

roberteg16 requested review from AndreasKaratzas and dllehr-amd as code owners September 8, 2026 10:12

roberteg16 requested review from eble-amd and removed request for AndreasKaratzas and dllehr-amd September 8, 2026 10:12

eble-amd approved these changes Sep 9, 2026

View reviewed changes

roberteg16 added this pull request to stack #1288 September 10, 2026 05:59

roberteg16 merged commit 17fcf0f into gfx11 Sep 10, 2026
8 of 9 checks passed

roberteg16 deleted the rogarcia.merge-upstream-118 branch October 12, 2026 12:56

Navigation Menu

Repository navigation

Upstream sync 22/N: merge 81e81dac1f (75 commits, conflict-free) - #1269

Upstream sync 22/N: merge 81e81dac1f (75 commits, conflict-free)#1269
roberteg16 merged 78 commits into
rogarcia.merge-upstream-117from
rogarcia.merge-upstream-118

roberteg16 commented Sep 7, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

20 participants

Conversation

roberteg16 commented Sep 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Test plan

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

20 participants

roberteg16 commented Sep 7, 2026 •

edited

Loading