xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-05-11 13:35:52 +08:00
Code Issues Packages Projects Releases Wiki Activity
12,369 Commits 155 Branches 89 Tags
Commit Graph

8660 Commits

Author SHA1 Message Date
yurekami
79e0db60ee Use --attention-backend flag instead of VLLM_ATTENTION_BACKEND env var 
Per reviewer feedback, the VLLM_ATTENTION_BACKEND environment variable
is being deprecated in favor of the --attention-backend CLI flag.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Signed-off-by: yurekami <yurekami@users.noreply.github.com>
 2025-12-24 15:16:10 +09:00
yurekami
3c8358c328 [v1][CP] Improve DCP/PCP/MTP error messages with actionable guidance 
Replace cryptic AssertionErrors with informative RuntimeErrors that:
- Explain what DCP (Decode Context Parallel) and PCP (Prefill Context
  Parallel) are
- List compatible attention backends
- Provide environment variable instructions (VLLM_ATTENTION_BACKEND)
- Include documentation links

Fixes #28407

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Signed-off-by: yurekami <249254018+yurekami@users.noreply.github.com>
 2025-12-18 20:05:16 +09:00
wangxiyuan
a85724bd6e
[Platform] Let EPD work with non-cuda platform (#30225) 
Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
 2025-12-18 06:45:29 +00:00
Yifan Qiao
11a89cf95c
[Fix][FlexAttention] return max logical block index to handle reused blocks (#30915) 
Signed-off-by: Yifan Qiao <yifanqiao@berkeley.edu>
 2025-12-18 06:42:21 +00:00
Li, Jiang
e3ab93c896
[CPU] Refactor CPU fused MOE (#30531) 
Signed-off-by: jiang1.li <jiang1.li@intel.com>
 2025-12-18 14:36:49 +08:00
Nathan Price
fc2ae6d617
fix: add warmup for audio preprocessing (#30706) 
Signed-off-by: Nathan Price <nathan@abridge.com>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
 2025-12-18 06:12:29 +00:00
Yihua Cheng
ec965569d9
[KV connector][LMCache] Only record the cuda event when there are request to store/load (#30814) 
Signed-off-by: ApostaC <yihua98@uchicago.edu>
 2025-12-18 05:31:34 +00:00
Vadim Gimpelson
717ac33d9c
[PERF] Qwen3-next. Add fp8 cutlass MoE tuned configs. chmod -x *MI308X.json (#29553) 
Signed-off-by: Vadim Gimpelson <vadim.gimpelson@gmail.com>
 2025-12-18 13:16:04 +08:00
zzhxxx
b166ef20e1
[refactor] Add prefix support to embed_tokens in DeepSeek MTP (#30788) 
Signed-off-by: zzhx1 <zzh_201018@outlook.com>
 2025-12-18 04:45:56 +00:00
Zhengxu Chen
5f2f3fba1d
[compile] Fix CI for test_gpt2_cache_hit (#30902) 
Signed-off-by: zhxchen17 <zhxchen17@fb.com>
 2025-12-17 20:22:23 -08:00
Matthew Bonanni
4a8412f773
[UX] Reduce DeepGEMM warmup log output to single progress bar (#30903) 
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
 2025-12-17 20:21:51 -08:00
Bowen Bao
0c738b58bc
[Quantization] Support Quark int4-fp8 w4a8 for MoE (#30071) 
Signed-off-by: Bowen Bao <bowenbao@amd.com>
 2025-12-18 04:20:42 +00:00
gnovack
5a3adf581e
fused_moe_lora PDL improvements (#30716) 
Signed-off-by: gnovack <gnovack@amazon.com>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
Co-authored-by: Jee Jee Li <pandaleefree@gmail.com>
 2025-12-17 19:55:00 -08:00
Isotr0py
6fe5887652
[Chore] Remove v0 dead code for Qwen2.5-omni (#30883) 
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
 2025-12-17 19:54:39 -08:00
Nicolò Lucchesi
bc3700e0cd
[NIXL] Support P tensor-parallel-size > D tensor-parallel-size (#27274) 
Signed-off-by: NickLucche <nlucches@redhat.com>
 2025-12-18 11:53:30 +08:00
Micah Williamson
fd8afdf38d
[ROCm][CI] Reduce Flakiness For test_async_scheduling Using ROCM_ATTN With FP32 (#30811) 
Signed-off-by: Micah Williamson <micah.williamson@amd.com>
 2025-12-18 10:27:37 +08:00
SungMinCho
a0b782f9cc
[Metrics] Model FLOPs Utilization estimation (#30738) 
Signed-off-by: SungMinCho <tjdals4565@gmail.com>
Signed-off-by: Mark McLoughlin <markmc@redhat.com>
Co-authored-by: Mark McLoughlin <markmc@redhat.com>
 2025-12-18 01:40:51 +00:00
Isotr0py
74a1ac38b0
[v1] Add PrefixLM support to TritonAttention backend (#30386) 2025-12-17 16:05:24 -08:00
Nathan Price
05a83dc6ee
feat(api): Eager chat template warmup to eliminate first-request latency (#30700) 
Signed-off-by: Nathan Price <nathan@abridge.com>
 2025-12-18 00:01:29 +00:00
Varun Sundar Rabindranath
e3fc374a9a
[BugFix] Workspace allocation during profile run : DeepEPHighThroughput + DeepGEMM (#30899) 2025-12-17 15:00:59 -08:00
Andrey Talman
e06d0bf0aa
2.9.1 PyTorch release update (#28495) 2025-12-17 12:20:22 -08:00
Matthew Bonanni
7eb6cb6c18
[Attention] Update tests to remove deprecated env vars (#30563) 
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
 2025-12-17 09:49:59 -08:00
Cyrus Leung
2497228ad4
[Chore] Factor out logic for requesting initial memory (#30868) 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
 2025-12-17 07:32:17 -08:00
KimHyemin
196cdc3224
[Model] Gemma3: Support untied word embeddings (#30827) 
Signed-off-by: www-spam <panmahm@naver.com>
 2025-12-17 07:11:18 -08:00
高鑫崧
b7b6a60aca
Adapt the old parameter enable_thinking in chat_template_kwargs (#30852) 
Signed-off-by: xinsong.gao <1418762819@qq.com>
Co-authored-by: Chauncey <chaunceyjiang@gmail.com>
 2025-12-17 07:10:59 -08:00
Jialin Ouyang
6e9dbcc50e
[Fix] uniform decode batch check (#30747) 
Signed-off-by: Jialin Ouyang <Jialin.Ouyang@gmail.com>
 2025-12-17 19:58:43 +08:00
Hank_
6482e3895b
chores: adjust the attn register param order (#30688) 
Signed-off-by: Hank <hcc.mayday@gmail.com>
 2025-12-17 19:58:16 +08:00
Harry Mellor
fb980eb2fd
Fix lazy import (#30858) 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
 2025-12-17 03:33:50 -08:00
baoqian426
84896fda22
[Bugfix] deepseek-V3.2 self.weights_proj has no bias (#30841) 
Signed-off-by: baoqian <1354987947@qq.com>
Signed-off-by: baoqian426 <1354987947@qq.com>
 2025-12-17 03:32:34 -08:00
Chauncey
9ad5b21710
[Refactor] [4/N] Move VLLM_SERVER_DEV endpoints into the serve directory (#30749) 
Signed-off-by: chaunceyjiang <chaunceyjiang@gmail.com>
 2025-12-17 02:27:30 -08:00
Wentao Ye
f284d7bd0c
[Bug] Fix AttributeError: 'ColumnParallelLinear' object has no attribute weight_scale_inv (#30823) 
Signed-off-by: yewentao256 <zhyanwentao@126.com>
 2025-12-17 02:00:35 -08:00
Zhengxu Chen
53cd7f868b
[compile] Recompile graph module during Dynamo cache loading. (#30743) 
Signed-off-by: Zhengxu Chen <zhxchen17@fb.com>
 2025-12-17 02:00:12 -08:00
danielafrimi
7b966ae2ba
[Fix]Load kv-cache dtype from hf_quant_config.json automatically (fix for reverted PR) (#30785) 
Signed-off-by: <>
Co-authored-by: root <root@gpu-937.slurm-workers-slurm.slurm.svc.cluster.local>
 2025-12-17 01:56:38 -08:00
Zhengxu Chen
9db1db5949
[compile] Ignore VLLM_FORCE_AOT_LOAD from cache factors (#30809) 
Signed-off-by: zhxchen17 <zhxchen17@fb.com>
 2025-12-17 01:56:24 -08:00
Zhengxu Chen
177c391db2
[compile] Disable aot when eager backend is used. (#30810) 
Signed-off-by: zhxchen17 <zhxchen17@fb.com>
 2025-12-17 01:55:56 -08:00
Michael Goin
519ef9a911
[UX] Make vllm bench serve discover model by default and use --input-len (#30816) 
Signed-off-by: mgoin <mgoin64@gmail.com>
 2025-12-17 01:55:30 -08:00
Ye (Charlotte) Qi
a100152288
[Kernels][FI] Skip trtllm attention when num_kv_heads=1 (#30842) 
Signed-off-by: Ye (Charlotte) Qi <yeq@meta.com>
 2025-12-17 01:54:21 -08:00
Andrew Xia
4c054d89aa
[Doc][ResponsesAPI] add documentation (#30840) 
Signed-off-by: Andrew Xia <axia@fb.com>
Co-authored-by: Andrew Xia <axia@fb.com>
 2025-12-17 01:53:02 -08:00
Xinyu Chen
3b1d440ede
CustomOp: grouped topk (#29575) 
Signed-off-by: Xinyu Chen <xinyu1.chen@intel.com>
 2025-12-17 17:43:00 +08:00
Asaf Joseph Gardin
a9e15c21ef
[Mamba] Removed disable cascade attn in MambaModelConfig (#30712) 
Signed-off-by: asafg <39553475+Josephasafg@users.noreply.github.com>
 2025-12-17 08:48:53 +00:00
Robin
20fda43151
[Bugfix][Frontend] Prevent IndexError in MiniMax M2 tool parser during streaming extraction (#30555) 
Signed-off-by: WangErXiao <863579016@qq.com>
 2025-12-17 16:37:57 +08:00
Yan Ma
4f735babb7
[XPU] fix broken fp8 online quantization for XPU platform (#30831) 
Signed-off-by: Yan Ma <yan.ma@intel.com>
 2025-12-17 00:28:13 -08:00
Li, Jiang
0cd5353644
[Bugfix][CPU] Fix CPU backend ROPE dispatch for VL models (#30829) 
Signed-off-by: jiang1.li <jiang1.li@intel.com>
Signed-off-by: Li, Jiang <bigpyj64@gmail.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
 2025-12-16 23:25:12 -08:00
Michael Goin
d4d2751732
Update note comment for flashinfer attention warmup (#30711) 
Signed-off-by: mgoin <mgoin64@gmail.com>
 2025-12-16 21:29:03 -08:00
Grzegorz K. Karch
f5db6385a1
Fix nemotron_nas intermediate_size computation (#30795) 
Signed-off-by: Grzegorz Karch <gkarch@nvidia.com>
 2025-12-17 01:06:28 +00:00
Nicolò Lucchesi
e087fbc393
[MM] Pass FA version in ViT Attn (#30756) 
Signed-off-by: NickLucche <nlucches@redhat.com>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
 2025-12-17 07:54:45 +08:00
TJian
2410132bb1
[ROCm] [Bugfix] Fix torch sdpa hallucination (#30789) 
Signed-off-by: tjtanaa <tunjian.tan@embeddedllm.com>
 2025-12-16 15:32:43 -08:00
Jinzhen Lin
ce96857fdd
[Kernel][Quantization][MoE] add marlin kernel support for turing (sm75) (#29901) 
Signed-off-by: Jinzhen Lin <jinzhen.ljz@antgroup.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
 2025-12-16 14:35:28 -08:00
Roger Wang
f5f51e5931
[Core][MM] Optimize encoder cache manager by operating with embeddings only (#30475) 
Signed-off-by: Roger Wang <hey@rogerw.io>
Co-authored-by: Sun Kim <sunytokki@gmail.com>
 2025-12-16 14:18:17 -08:00
Lucas Wilkinson
9fec0e13d5
[Attention] Cache attention metadata builds across hybrid KV-cache groups (#29627) 
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Co-authored-by: Stanislaw Wozniak <stw@zurich.ibm.com>
 2025-12-16 17:10:16 -05:00