Fix: Correct max_model_len derivation from config.json for Mistral format#17777

Closed

princepride wants to merge 72 commits intovllm-project:mainfrom

princepride:patch-1

Commits on May 10, 2025

Fix https://github.com/vllm-project/vllm/issues/17747 Bug
princepride
committed
Update config.py
princepride
committed
Update config.py
princepride
committed
Update config.py
princepride
committed
Update config.py
princepride
committed
[doc] update the issue link (#17782)

reidliu41
authored and
princepride
committed
[ROCm][FP8][Kernel] FP8 quantization fused into Custom Paged Attention (#17139)

gshtras
authored and
princepride
committed
Only depend on importlib-metadata for Python < 3.10 (#17776)

tiran
authored and
princepride
committed
[Bugfix] Fix Video IO error for short video (#17791)

Isotr0py
authored and
princepride
committed
Fix and simplify `deprecated=True` CLI `kwarg` (#17781)

hmellor
authored and
princepride
committed
[Bugfix] Fix missing lora name mapping for lora without prefix (#17793)

Isotr0py
authored and
princepride
committed
[Quantization] Quark MXFP4 format loading (#16943)

BowenBao
authored and
princepride
committed
[Hardware][TPU][V1] Multi-LoRA implementation for the V1 TPU backend (#14238)

authored and
princepride
committed
[BugFix] Avoid secondary missing `MultiprocExecutor.workers` error (#17811)

njhill
authored and
princepride
committed
[Core][Feature] Input metadata dump on crash (#13407)

wallashss
authored and
princepride
committed
[Chore][Doc] uses model id determined from OpenAI client (#17815)

aarnphm
authored and
princepride
committed
Don't call the venv `vllm` (#17810)

hmellor
authored and
princepride
committed
[BugFix] Fix `--disable-log-stats` in V1 server mode (#17600)

njhill
authored and
princepride
committed
[Core] Support full cuda graph in v1 (#16072)

authored and
princepride
committed
Improve exception reporting in MP engine (#17800)

vmarkovtsev
authored and
princepride
committed
[Installation] OpenTelemetry version update (#17771)

Xarbirus
authored and
princepride
committed
Only log non-default CLI args for online serving (#17803)

hmellor
authored and
princepride
committed
[V1] Add VLLM_ALLOW_INSECURE_SERIALIZATION env var (#17490)

authored and
princepride
committed
[Kernel][Hardware][AMD] Bf16 mfma opt for ROCm skinny GEMMs (#17071)

authored and
princepride
committed
[Hardware][Power] Enable compressed tensor W8A8 INT8 quantization for POWER (#17153)

authored and
princepride
committed
[Hardware][Intel-Gaudi] Support Automatic Prefix Caching on HPU (#17648)

adobrzyn
authored and
princepride
committed
[Frontend] Chat template fallbacks for multimodal models (#17805)

DarkLight1337
authored and
princepride
committed
[Qwen3]add qwen3-235b-bf16 fused moe config on A100 (#17715)

Ximingwang-09
authored and
princepride
committed
[Bugfix] Fix bad words for Mistral models (#17753)

qionghuang6
authored and
princepride
committed
[Misc] support model prefix & add deepseek vl2 tiny fused moe config (#17763)

xsank
authored and
princepride
committed
[Bugfix] Fix tool call template validation for Mistral models (#17644)

authored and
princepride
committed
[TPU] Fix the test_sampler (#17820)

bythew3i
authored and
princepride
committed
[Bugfix] Fix quark fp8 format loading on AMD GPUs (#12612)

authored and
princepride
committed
[Doc] Fix a typo in the file name (#17836)

DarkLight1337
authored and
princepride
committed
[Easy] Eliminate c10::optional usage in vllm/csrc (#17819)

houseroad
authored and
princepride
committed
[Misc] add chatbox integration (#17828)

reidliu41
authored and
princepride
committed
Fix transient dependency error in docs build (#17848)

hmellor
authored and
princepride
committed
[Bugfix] `use_fast` failing to be propagated to Qwen2-VL image processor (#17838)

DarkLight1337
authored and
princepride
committed
[Misc] Delete LoRA-related redundancy code (#17841)

jeejeelee
authored and
princepride
committed
[CI] Fix test_collective_rpc (#17858)

russellb
authored and
princepride
committed
[V1] Improve VLLM_ALLOW_INSECURE_SERIALIZATION logging (#17860)

russellb
authored and
princepride
committed
[Test] Attempt all TPU V1 tests, even if some of them fail. (#17334)

yarongmu-google
authored and
princepride
committed
[CI] Prune down lm-eval small tests (#17012)

mgoin
authored and
princepride
committed
Fix noisy warning for uncalibrated q_scale/p_scale (#17414)

mgoin
authored and
princepride
committed
Add cutlass support for blackwell fp8 blockwise gemm (#14383)

wenscarl
authored and
princepride
committed
[FEAT][ROCm]: Support AITER MLA on V1 Engine (#17523)

authored and
princepride
committed
[V1][Structured Output] Update llguidance (`>= 0.7.11`) to avoid AttributeError (no `StructTag`) (#17839)

shen-shanshan
authored and
princepride
committed
[Attention] MLA move rotary embedding to cuda-graph region (#17668)

LucasWilkinson
authored and
princepride
committed
[BUGFIX]: return fast when request requires prompt logprobs (#17251)

andyxning
authored and
princepride
committed
[Docs] Add Slides from NYC Meetup (#17879)

simon-mo
authored and
princepride
committed
[Doc] Update several links in reasoning_outputs.md (#17846)

windsonsea
authored and
princepride
committed
[Doc] remove visible token in doc (#17884)

yma11
authored and
princepride
committed
[Bugfix][ROCm] Fix AITER MLA V1 (#17880)

vllmellm
authored and
princepride
committed
[Bugfix][CPU] Fix broken AVX2 CPU TP support (#17252)

Isotr0py
authored and
princepride
committed
Fix Whisper crash caused by invalid``` max_num_batched_tokens``` config (#17853)

inkcherry
authored and
princepride
committed
Change `top_k` to be disabled with `0` (still accept `-1` for now) (#17773)

hmellor
authored and
princepride
committed
[Misc] add dify integration (#17895)

reidliu41
authored and
princepride
committed
[BugFix][AMD] Compatible patch for latest AITER(05/07/2025) (#17864)

qli88
authored and
princepride
committed
[v1] Move block management logic from KVCacheManager to SpecializedManager (#17474)

heheda12345
authored and
princepride
committed
[CI/Build] Automatically retry flaky tests (#17856)

DarkLight1337
authored and
princepride
committed
Revert "[BugFix][AMD] Compatible patch for latest AITER(05/07/2025)" (#17910)

mgoin
authored and
princepride
committed
[Misc] Add references in ray_serve_deepseek example (#17907)

ruisearch42
authored and
princepride
committed
[Misc] Auto fallback to float16 for pre-Ampere GPUs when detected bfloat16 config (#17265)

Isotr0py
authored and
princepride
committed
Update CT WNA16MarlinMoE integration (#16666)

mgoin
authored and
princepride
committed
Handle error when `str` passed to `/v1/audio/transcriptions` (#17909)

hmellor
authored and
princepride
committed
Add option to use torch._inductor.standalone_compile (#17057)

zou3519
authored and
princepride
committed
[V1][Spec Decoding] Include bonus tokens in mean acceptance length (#17908)

markmc
authored and
princepride
committed
Improve configs - the rest! (#17562)

hmellor
authored and
princepride
committed
AMD conditional all test execution // new test groups (#17556)

Alexei-V-Ivanov-AMD
authored and
princepride
committed
[Hardware/NVIDIA/Kernel] Enable nvidia/DeepSeek-R1-FP4 Model (#16362)

pavanimajety
authored and
princepride
committed
[V1][Spec Decoding] Log accumulated metrics after system goes idle (#17913)

markmc
authored and
princepride
committed
adjust the code
princepride
committed