Skip to content

Merge upstream→gfx11#1038

Merged
eble-amd merged 16 commits into
ROCm:gfx11from
eble-amd:merge-from-upstream
Jun 30, 2026
Merged

Merge upstream→gfx11#1038
eble-amd merged 16 commits into
ROCm:gfx11from
eble-amd:merge-from-upstream

Conversation

@eble-amd

@eble-amd eble-amd commented Jun 30, 2026

Copy link
Copy Markdown

No conflicts.

zvik and others added 16 commits June 4, 2026 14:47
@zvik @claude
Add model support for granite speech plus (vllm-project#43519)
b21443e 
Signed-off-by: Zvi Kons[WSL] <zvi@il.ibm.com>
Signed-off-by: Zvi Kons (BlueVela) <zvi@il.ibm.com>
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
@tc-mb
[Bugfix] MiniCPM-V-4.6 video inference crash: placeholder count misma…
3dbb4e0 
…tches visual embedding count (vllm-project#44509)

Signed-off-by: tc-mb <tianchi_cai@icloud.com>
@tlrmchlsmth
[Core] Freeze garbage collector in workers after model initialization (…
4cc78c9 
…vllm-project#44363)

Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
@jwzheng96 @yewentao256
[Bugfix] Reject non-positive values for ParallelConfig int knobs (vll…
99ef652 
…m-project#44057)

Signed-off-by: jwzheng96 <jianweizheng@pku.edu.cn>
Signed-off-by: JianweiZheng <32029023+jwzheng96@users.noreply.github.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
@divakar-amd @AndreasKaratzas
[ROCm][CI] Add test for Aiter unified attn kernel (vllm-project#44436)
06f9463 
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
Signed-off-by: Divakar Verma <divakar.verma@amd.com>
Co-authored-by: Andreas Karatzas <akaratza@amd.com>
@fadara01
[DOC] Add INT8 W4A8 docs and Arm's supported quantization schemes (vl…
3da29aa 
…lm-project#34894)

Signed-off-by: Fadi Arafeh <fadi.arafeh@arm.com>
@taneem-ibrahim
[Misc] Add unit tests for pooler head classes (vllm-project#44471)
8d9536a 
Signed-off-by: Taneem Ibrahim <taneem.ibrahim@gmail.com>
@bnellnm
[Bugfix] Fix test_cutlass_moe.py (vllm-project#44380)
439203d 
Signed-off-by: Bill Nell <bnell@redhat.com>
@adobrzyn
[Kernel][Test] Extend lightning_attn and awq_triton kernel tests to X…
a947f7a 
…PU (vllm-project#43307)

Signed-off-by: Dobrzyniewicz, Agata <agata.dobrzyniewicz@intel.com>
@tushar00jain
use split_group for pytorch process group creation (vllm-project#41980)
38fd240 
Signed-off-by: Tushar Jain <tushar00jain@users.noreply.github.com>
Co-authored-by: Tushar Jain <tushar00jain@users.noreply.github.com>
@yewentao256
[Logs Refactor] Optimize shutdown logs, easier to follow and consiste…
41a4829 
…nt (vllm-project#43707)

Signed-off-by: yewentao256 <zhyanwentao@126.com>
@ZJY0516
[mamba] unify KDA conv states into one cache to match 2-state SSM lay…
a55fccf 
…out (vllm-project#44539)
@galletas1712
fix: keep DeepSeek V4 RoPE cache on inv_freq device (vllm-project#43926)
b7c5baf 
Signed-off-by: Schwinn Saereesitthipitak <schwinns@nvidia.com>
Signed-off-by: Schwinn Saereesitthipitak <17022745+galletas1712@users.noreply.github.com>
@BugenZhao
[Rust Frontend] Skip loading multimodal processor if `--language-mode…
62d6f06 
…l-only` is specified (vllm-project#44500)

Signed-off-by: Bugen Zhao <i@bugenzhao.com>
@zufangzhu
[XPU][MoE] support block_fp8_moe on xpu (vllm-project#42139)
063ce98 
Signed-off-by: Zhu, Zufang <zufang.zhu@intel.com>
Signed-off-by: zofia <110436990+zufangzhu@users.noreply.github.com>
@eble-amd
Merge commit '063ce98fb7104dba93f72d56c094fb8b708bd793' into merge-fr…
e7e0ee0 
…om-upstream
@eble-amd

eble-amd commented Jun 30, 2026

Copy link
Copy Markdown
Author

Results of manual sanity test on gfx1151:

Arguments: --model Qwen/Qwen2.5-0.5B-Instruct-AWQ --num-prompts 10 --max-model-len 4096 --kill-e
xisting-vllm-processes --input-len 3968 --output-len 128 --target-gpu-memory-gb 10 --max-num-seq
s 1 -e TORCH_ROCM_AOTRITON_ENABLE_EXPERIMENTAL=1 -e FLASH_ATTENTION_TRITON_AMD_ENABLE=TRUE -e TO
RCH_BLAS_PREFER_HIPBLASLT=1
Prefill 23356.73 tokens/s; TTFT 170 ms
Decode 312.3 tokens/s (TPOT 3.20 ms)
End-to-end latency 576 ms (median)
Average CPU utilization 6%

@eble-amd

eble-amd commented Jun 30, 2026 •
edited
Loading

Copy link
Copy Markdown
Author

Results of performance regression tests on gfx1151:

python -m pytest -v tests/kernels/quantization/test_hybrid_w4a16_perf.py
...
========================= 128 passed, 17 warnings in 530.51s (0:08:50) =========================
...

@eble-amd

eble-amd commented Jun 30, 2026

Copy link
Copy Markdown
Author

The build failed due to No space left on device. I'm retriggering to see whether it is reproducible.

@eble-amd eble-amd marked this pull request as ready for review June 30, 2026 19:31
@eble-amd eble-amd requested a review from mgehre-amd June 30, 2026 19:31
@eble-amd eble-amd merged commit f8bfb4d into ROCm:gfx11 Jun 30, 2026
10 of 11 checks passed
@eble-amd eble-amd deleted the merge-from-upstream branch June 30, 2026 19:52
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@mgehre-amd mgehre-amd mgehre-amd approved these changes
@AndreasKaratzas AndreasKaratzas Awaiting requested review from AndreasKaratzas AndreasKaratzas is a code owner

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.