New PR number #25693#25688

Closed

charlifu wants to merge 275 commits intovllm-project:mainfrom

ROCm:amd/aiter_fusion_pass

This pull request is big! We're only showing the most recent 250 commits

Commits on Sep 25, 2025

[Rocm] [quantization] Fix quark ptpc moe and add test case (#24649)

authored and
charlifu
committed
Add more documentation and improve usability of lognormal dist (benchmark_serving_multi_turn) (#23255)

pliops-daniels
authored and
charlifu
committed
[XPU] Fix xpu model runner call torch.cuda APIs (#25011)

jikunshang
authored and
charlifu
committed
[EPLB] Support EPLB for Mixtral Model (#22842)

authored and
charlifu
committed
[Core][MultiModalHasher] Hash images without converting image mode (#24969)

lgeiger
authored and
charlifu
committed
[Model] Pass param prefix to LLMHead (#24862)

whx-sjtu
authored and
charlifu
committed
[Model] Apply SharedFusedMoE to glm4_moe. (#24849)

whx-sjtu
authored and
charlifu
committed
[Core] Remove tokenizer group in vLLM (#24078)

zhuohan123
authored and
charlifu
committed
[Docs] Fix griffe warning in base_static_graph.py (#25018)

windsonsea
authored and
charlifu
committed
[DP] Create placement groups by ray_device_key (#25026)

authored and
charlifu
committed
[Frontend] Support returning all prompt logprobs (#24956)

chaunceyjiang
authored and
charlifu
committed
[BugFix] enable DOTALL to match multi-line tool_call parameters in extract_tool_call_required_streaming (#24668)

shijun-yin
authored and
charlifu
committed
[Misc] Avoid use of deprecated `AutoModelForVision2Seq` (#25065)

DarkLight1337
authored and
charlifu
committed
Add RADIO Vision Encoder Support to vLLM (#24595)

authored and
charlifu
committed
[Bugfix] Fix Stream usage in CPU model runner and OneDNN kernel check (#25046)

bigPYJ1151
authored and
charlifu
committed
Apply fixes for CUDA 13 (#24599)

Aidyn-A
authored and
charlifu
committed
[fix] lora benchmarks pass no_lora_flag_cpu (#23774)

authored and
charlifu
committed
[Bugfix][Qwen3-Next] fixes the varlen issue in qwen3-next's MTP implementation. (#24957)

sighingnow
authored and
charlifu
committed
[Docs] improve code formatting and comments for eliminate griffe build warning. (#25010)

samzong
authored and
charlifu
committed
Remove old cutlass mla (#23961)

MatthewBonanni
authored and
charlifu
committed
[Docs] vllm/benchmarks/datasets.py fix docstring param format. (#24970)

samzong
authored and
charlifu
committed
[CI Bugfix] Fix failing test_invalid_env (#25078)

mgoin
authored and
charlifu
committed
[V0 Deprecation] Remove V0 Core tests (#25082)

WoosukKwon
authored and
charlifu
committed
cleanup: remove adapter commons (#25045)

authored and
charlifu
committed
Remove unused find_cuda_init helper script (#25044)

simon-mo
authored and
charlifu
committed
[V0 Deprecation] Remove unused output processor util (#25023)

WoosukKwon
authored and
charlifu
committed
Change log level from info to debug for IOProcessor (#24999)

mgoin
authored and
charlifu
committed
[CI] Revert back prepare_prompts and check_answers (#25087)

WoosukKwon
authored and
charlifu
committed
[V0 Deprecation] Remove V0 tests in test_sequence.py (#25088)

WoosukKwon
authored and
charlifu
committed
[CI Bugfix] Fix failing test_model_load_with_params tests due to tokenizer refactor (#25086)

mgoin
authored and
charlifu
committed
[V1] Logits processor docs (#22919)

authored and
charlifu
committed
[Misc] Update owners for KV connector and V1 offloading (#25041)

ApostaC
authored and
charlifu
committed
[Bugfix] Update import path for bc_linter_include (#24766)

mmangkad
authored and
charlifu
committed
[BUG] Exclude .pth files when pulling remote files (#25092)

ahao-anyscale
authored and
charlifu
committed
[Kernel] Faster pre-processing time for W4A8 (#23972)

czhu-cohere
authored and
charlifu
committed
[gpt-oss][2] fix types for streaming (#24556)

qandrew
authored and
charlifu
committed
[Bugfix][B200] Fix `cutlass_mla` hang (#24966)

authored and
charlifu
committed
Aiter mha fp8 fix (#24991)

authored and
charlifu
committed
Disable failing GPT-OSS Eval (Blackwell) for now (#25107)

mgoin
authored and
charlifu
committed
[Bugfix] Refactor Flashinfer TRTLLM attention kernel selection logic (#24600)

authored and
charlifu
committed
Add a batched auto tune script (#25076)

authored and
charlifu
committed
[Bugfix] Fix accuracy issue for silu_mul + nvfp4 quant fusion kernel (#24833)

authored and
charlifu
committed
[Kernel] Delegate construction of FusedMoEQuantConfig to FusedMoEMethodBase subclasses (#22537)

bnellnm
authored and
charlifu
committed
[V0 Deprecation] Remove V0 Engine tests (#25114)

WoosukKwon
authored and
charlifu
committed
[V0 Deprecation] Remove V0 Tracing & Metrics tests (#25115)

WoosukKwon
authored and
charlifu
committed
[V0 Deprecation] Remove misc V0 tests (#25118)

WoosukKwon
authored and
charlifu
committed
[V0 Deprecation] Skip PP test (#25128)

WoosukKwon
authored and
charlifu
committed
[Kernels] Enable DeepGEMM by default (#24462)

bnellnm
authored and
charlifu
committed
[MM Encoder] Apply DP ViT for Qwen3-VL model series (#24955)

authored and
charlifu
committed
[Docs] Clean up the contributing README (#25099)

hmellor
authored and
charlifu
committed
[Core][MM] Cleanup `MultiModalCache` (#25006)

lgeiger
authored and
charlifu
committed
[Bugfix][Qwen3-Next] add prefixes to shared_expert in qwen3-next and mlp in qwen2moe to successfully load ignored params in quantized models (#24960)

authored and
charlifu
committed
[Kernels] Overlap shared experts with combine instead of dispatch (#24254)

bnellnm
authored and
charlifu
committed
[Model] enable data parallel for InternVL vision encoder (#23909)

authored and
charlifu
committed
Mark prompt logprobs as incompatible with prompt embeds at API level (#25077)

qthequartermasterman
authored and
charlifu
committed
[XPU] Whisper model support on XPU Platform (#25123)

chaojun-zhang
authored and
charlifu
committed
[EPLB] Add EPLB support for hunyuan_v1 (#23078)

666even666
authored and
charlifu
committed
[V0 Deprecation] Remove more V0 tests (#25117)

WoosukKwon
authored and
charlifu
committed
[Spec Decode] Efficient padded speculation (#24539)

benchislett
authored and
charlifu
committed
[benchmark] add peak throughput metrics and plot (#23867)

simon-mo
authored and
charlifu
committed
[CLI] Use streaming in CLI chat and completion commands (#23769)

simon-mo
authored and
charlifu
committed
[Kernel] Better inf handling for grouped topk cu (#24886)

lumina37
authored and
charlifu
committed
[Docs] Fix API Reference (#25140)

hmellor
authored and
charlifu
committed
Retrieve `sliding_window` from text config in Gemma3 MM (#25085)

hmellor
authored and
charlifu
committed
[Bugfix] when use s3 model cannot use default load_format (#24435)

authored and
charlifu
committed
[Qwen] Add fp8 checkpoint support for qwen3-next. (#25079)

sighingnow
authored and
charlifu
committed
Add 'path' option to ImagePrompt data_format (#25081)

gfinol
authored and
charlifu
committed
[Doc] Fix cross-reference warnings (#25058)

authored and
charlifu
committed
[Chore] Cleanup guided namespace, move to structured outputs config (#22772)

authored and
charlifu
committed
Fix: Add explicit #include <omp.h> for OpenMP compatibility on certain toolchains (#24951)

ihb2032
authored and
charlifu
committed
silu-v1: Fix EPS not being used during max-reduction (#25069)

elvircrn
authored and
charlifu
committed
[Frontend] Support setting logprobs to -1 (#25031)

chaunceyjiang
authored and
charlifu
committed
[Model] Improve Pooling Model (#25149)

jeejeelee
authored and
charlifu
committed
Move `StructuredOutputsConfig` from `config/__init__.py` to `config/structured_outputs.py` (#25153)

hmellor
authored and
charlifu
committed
[Docs] Fix pooling-params doc references in openai_compatible_server.md (#24939)

yankay
authored and
charlifu
committed
[Docs] add the parallel sampling usage in LLMEngine and AsyncLLM (#24222)

gigit0000
authored and
charlifu
committed
Fix forward reference warning in documentation (#25150)

hmellor
authored and
charlifu
committed
Fix `validate-config` pre-commit check (#25157)

hmellor
authored and
charlifu
committed
[Bugfix][Mamba] - Fix Conv State Kernel FP32 Support (#24883)

Josephasafg
authored and
charlifu
committed
[Misc] Clean up flags in `vllm bench serve` (#25138)

ywang96
authored and
charlifu
committed
[Structured Output][Refactor] Move `apply_grammar_bitmask()` method from `ModelRunner` to structured output utils (#21999)

shen-shanshan
authored and
charlifu
committed
Refactor dense FP8 tensor/channel/block utils and add CT FP8 block (#21404)

mgoin
authored and
charlifu
committed
[Misc] Add kv-connector label (#25156)

NickLucche
authored and
charlifu
committed
[Kernel] Decouple Tile Size from Block Size in Triton Unified Attention Kernel (#21197)

jvlunteren
authored and
charlifu
committed
[PERF] Add `conv1d` metadata to GDN attn (#25105)

vadiklyutiy
authored and
charlifu
committed
feat(api): Return 503 on /health when engine is dead (#24897)

authored and
charlifu
committed
[New Model] Support BertForTokenClassification / Named Entity Recognition (NER) task (#24872)

authored and
charlifu
committed
[Docs] Fix warnings in mkdocs build (continued) (#25163)

Zerohertz
authored and
charlifu
committed
Enable Allgather/ReduceScatter backend for NaiveAllToAll (#23964)

authored and
charlifu
committed
[Misc] Add codeowner for Transformers backend (#25180)

hmellor
authored and
charlifu
committed
[spec decode] Fix MTP inference path for MiMo-7B model (#25136)

authored and
charlifu
committed
[ROCm][CI/Build] Use ROCm7.0 as the base (#25178)

gshtras
authored and
charlifu
committed
fix aiter fp8 linear support
charlifu
committed
[ROCm][AITER][Bugfix] Switch AITER to use PIECEWISE_AND_FULL compilation (#25104)

Rohan138
authored and
charlifu
committed
[KV offload][1/N] Introduce an offloading component (#19848)

orozery
authored and
charlifu
committed
[V0 Deprecation] Remove AsyncLLMEngine (#25025)

WoosukKwon
authored and
charlifu
committed
[fix]: remove data type hardcoding from gptoss model implementation (#23807)

nikhil-arm
authored and
charlifu
committed
[feat]: Create interface for model-specific M-RoPE (#24194)

authored and
charlifu
committed
[Bug] Fix `returned_lse` not Defined issue (#25106)

authored and
charlifu
committed
[Bug] Fix torch Compilation Cache Hit Error (#25093)

yewentao256
authored and
charlifu
committed
[V0 Deprecation] Remove unused async_timeout.py (#25190)

WoosukKwon
authored and
charlifu
committed
[KV offload][1b/N] rename offloading to kv_offload (#25191)

orozery
authored and
charlifu
committed
[BugFix] Fix DeepGEMM warmup, no m.weight_scale_inv (#25206)

LucasWilkinson
authored and
charlifu
committed
[CORE] Prompt Embeddings Support for v1 Engine (#24278)

authored and
charlifu
committed
[KV offload][2/N] Introduce LRU-based CPU offloading management (#20075)

orozery
authored and
charlifu
committed
[gpt-oss] Add ResponseReasoningPartAddedEvent, ResponseReasoningPartDoneEvent for streaming (#24938)

qandrew
authored and
charlifu
committed
[Perf] Optimize memory peak during EAGLE model loading. (#24585)

candyzone
authored and
charlifu
committed
[Misc] Clean up MM profiling warnings (#25222)

ywang96
authored and
charlifu
committed
[Docs] Fix griffe warnings in vllm/multimodal (#25216)

windsonsea
authored and
charlifu
committed
[OOT] Support sync_model_loading for OOT (#25126)

xuechendi
authored and
charlifu
committed
[Build] Update Xgrammar to 0.1.24 to get a CVE fix (#25188)

russellb
authored and
charlifu
committed
[CPU] Disable oneDNN linear on non-x86 platforms (#25166)

bigPYJ1151
authored and
charlifu
committed
[Bugfix][CPU] Add placeholder to avoid import errors when using fused_moe ops on platforms without triton (#25137)

bigPYJ1151
authored and
charlifu
committed
[Misc] Cleanup test conftest for deprecated encoder-decoder models (#25231)

Isotr0py
authored and
charlifu
committed
[bugfix] fix MHA for models like OpenGVLab/InternVL3_5-38B (#25146)

authored and
charlifu
committed
[Kernel][Performance] Add Triton kernel for Qwen3-VL interleaved MRoPE (#25055)

Isotr0py
authored and
charlifu
committed
[Bugfix][Perf] Misc fixes for Qwen3 VL (#25238)

ywang96
authored and
charlifu
committed
Move `PoolerConfig` from `config/__init__.py` to `config/pooler.py` (#25181)

hmellor
authored and
charlifu
committed
[P/D][Nixl] Introduce `KVTransferMetrics` and aggregation strategy (#22188)

NickLucche
authored and
charlifu
committed
[V0 Deprecation] Remove V0 logic from `get_input_embeddings` interface (#25242)

DarkLight1337
authored and
charlifu
committed
[Qwen] Remove cuda hard-code in qwen3 next (#25243)

wxsIcey
authored and
charlifu
committed
Update CODEOWNERS (#25269)

hmellor
authored and
charlifu
committed
Move `ModelConfig` from `config/__init__.py` to `config/model.py` (#25252)

hmellor
authored and
charlifu
committed
refactor(benchmarks): add type annotations to wait_for_endpoint parameters (#25218)

samzong
authored and
charlifu
committed
[KV offload][3/N] Add worker-side CPU support (#21448)

orozery
authored and
charlifu
committed
[Frontend] Pass API server count to each process (#23717)

DarkLight1337
authored and
charlifu
committed
[Core] Modify the initialization parameters of the lora manager (#25249)

jeejeelee
authored and
charlifu
committed
Remove Redundant Assignment in Qwen3_VisionPatchMerger (#25224)

authored and
charlifu
committed
Encoder model support for the Transformers backend (#25174)

hmellor
authored and
charlifu
committed
[CI/Build] fix test function_calling (#25072)

chaunceyjiang
authored and
charlifu
committed
[Core][Prefix Hash] Fix prefix hash metrics sliding window maintainance (#24990)

Jialin
authored and
charlifu
committed
[Docs] add __init__.py to vllm/model_executor/layers/quantization/compressed_tensors/transform (#24974)

samzong
authored and
charlifu
committed
[bugfix] fix structured outputs key missing issue from #24929 (#25195)

luccafong
authored and
charlifu
committed
[KV offload][4/N] Offloading KV connector (#22595)

orozery
authored and
charlifu
committed
Optimize triton unified attention performance for sliding window attention (#24390)

zixi-qi
authored and
charlifu
committed
[Bugfix] GPT OSS Attritbute error on H100 (#25228)

authored and
charlifu
committed
[Bugfix] Fix chunked a2_scales in modular kernels (#25264)

bnellnm
authored and
charlifu
committed
Specify platform in `pip-compile` `pre-commit` hook so it runs on MacOS (#25273)

hmellor
authored and
charlifu
committed
[Perf] Use FlashInfer RoPE for RotaryEmbedding.forward_cuda when available (#21126)

authored and
charlifu
committed
[BugFix] Make FlashInferMetadataBuilder non-blocking (#25040)

authored and
charlifu
committed
Fix: Correct FusedMoE layer reference in auto_round quantization (#24818)

authored and
charlifu
committed
[Frontend] Responses API messages out, just harmony for now (#24985)

authored and
charlifu
committed
[Compile] Fix Compile Warning for Ignoring `MIN_BLOCK_PER_SM` (#25193)

yewentao256
authored and
charlifu
committed
Enable modelopt gemma3 nvfp4/fp8, make workflow more robust (#22771)

authored and
charlifu
committed
allow disable flashinfer prefill (#25276)

luccafong
authored and
charlifu
committed
[BugFix] Fix async scheduling CPU tensor race take 2 (#25279)

njhill
authored and
charlifu
committed
[Bugfix] Remove VLLM_TEST_DYNAMO_FULLGRAPH_CAPTURE #2969 (#25090)

Lucaskabela
authored and
charlifu
committed
Don't skip special tokens with hermes-style tool calling (#25281)

maxdebayser
authored and
charlifu
committed
test: Remove vestigial skip for prompt embeds tests after landing v1 Prompt Embeds support (#25291)

qthequartermasterman
authored and
charlifu
committed
[docs] Prompt Embedding feature support (#25288)

qthequartermasterman
authored and
charlifu
committed
[torch.compile] CUDAGraph Inductor partition integration (#24281)

authored and
charlifu
committed
[BugFix] Ensure appropriate guards in destructors (#25284)

authored and
charlifu
committed
[Misc] Support more collective_rpc return types (#25294)

njhill
authored and
charlifu
committed
Improve weight loading for encoder models in Transformers backend (#25289)

hmellor
authored and
charlifu
committed
[BUGFIX] GPTQ quantization compatibility for Qwen3 Next MOE models (AutoGPTQ and AutoRound-GPTQ) (#25268)

JartX
authored and
charlifu
committed
[BugFix] Exclude self when checking for port collision (#25286)

njhill
authored and
charlifu
committed
[BUG FIX][NON-CUDA]quick fix to avoid call cudagraph_unsafe in attention (#25298)

xuechendi
authored and
charlifu
committed
[Bugfix] fix tool call arguments is empty (#25223)

authored and
charlifu
committed
[Optimization] Avoid repeated model architecture conversion for pooling models (#25261)

DarkLight1337
authored and
charlifu
committed
[Hybrid Allocator] Support full attention with different hidden size (#25101)

heheda12345
authored and
charlifu
committed
[Bugfix] Fix Qwen3-VL-MoE weight loading for EP (#25300)

ywang96
authored and
charlifu
committed
[V1] Support `LLM.apply_model` (#18465)

DarkLight1337
authored and
charlifu
committed
[CI Failure] Disable FlashInfer RoPE to unblock CI (#25299)

mgoin
authored and
charlifu
committed
[Docs] Fix warnings in mkdocs build (continued) (#25042)

wwl2755
authored and
charlifu
committed
Generate _ModelInfo properties file when loading to improve loading speed (#23558)

authored and
charlifu
committed
[Model] Cleanup InternViT's data parallel implementation (#25306)

Isotr0py
authored and
charlifu
committed
[Core] Enable sharded state loader for V1 engine and enhance test coverage (#25308)

lirong-lirong
authored and
charlifu
committed
[V0 Deprecation] Enable the remaining multimodal tests in V1 (#25307)

DarkLight1337
authored and
charlifu
committed
[Docs] Fix warnings in vllm/profiler and vllm/transformers_utils (#25220)

windsonsea
authored and
charlifu
committed
[V0 Deprecation] Remove LLMEngine (#25033)

WoosukKwon
authored and
charlifu
committed
[V0 Deprecation] Remove V0 Output Processor (#25320)

WoosukKwon
authored and
charlifu
committed
[Chore] Remove unused sampler in models (#25324)

WoosukKwon
authored and
charlifu
committed
[CI] Skip tests failing on main (#25326)

WoosukKwon
authored and
charlifu
committed
[V0 Deprecation] Remove V0 core (#25321)

WoosukKwon
authored and
charlifu
committed
[Doc] improve test-pipeline.yaml documentation (#25305)

authored and
charlifu
committed
[V0 Deprecation] Remove V0 model runner base & simplify worker base (#25328)

WoosukKwon
authored and
charlifu
committed
[Multi Modal][Performance] Fused Q,K's apply_rope in more models (#25005)

authored and
charlifu
committed
[V0 Deprecation] Remove from_seq_group methods (#25330)

WoosukKwon
authored and
charlifu
committed
[V0 Deprecation] Remove V0 MP executor (#25329)

WoosukKwon
authored and
charlifu
committed
[V1] Add sliding window support to Flex Attention backend (#24089)

Isotr0py
authored and
charlifu
committed
[MM][Perf] Minor Optimization on Qwen3-VL `fast_pos_embed_interpolate` (#25337)

ywang96
authored and
charlifu
committed
[Bugfix] Typos in error message for missing model config file (#25339)

simondanielsson
authored and
charlifu
committed
[Optimization] Cache chat template result when processor fails to be loaded (#25341)

DarkLight1337
authored and
charlifu
committed
[V0 Deprecation] Remove V0 Sequence class & Sampler (#25332)

WoosukKwon
authored and
charlifu
committed
[V0 Deprecation] Remove async_output_proc, preemption mode, delay factor (#25334)

WoosukKwon
authored and
charlifu
committed
feat: Enable engine-level arguments with speculators models (#25250)

authored and
charlifu
committed
[V0 Deprecation] Remove V0 sampling metadata (#25345)

WoosukKwon
authored and
charlifu
committed
[Perf] Further optimization for Qwen3-VL `fast_pos_embed_interpolate` (#25347)

Isotr0py
authored and
charlifu
committed
Remove V0 attention backends (#25351)

WoosukKwon
authored and
charlifu
committed
[Bugfix][V0 Deprecation][CI] use async mock and await for async method (#25325)

KKSK-DON
authored and
charlifu
committed
Multimodal - audio tests (#25285)

debroy-rh
authored and
charlifu
committed
[Model] Support Dots OCR (#24645)

authored and
charlifu
committed
[Docs] GSM8K Accuracy Evaluation doc update (#25360)

david6666666
authored and
charlifu
committed
[Bugfix] Fix hermes tool parser handling of non-string argument types (#22002)

authored and
charlifu
committed
[V0 Deprecation] Remove V0-only methods in multi-modal registry (#25362)

DarkLight1337
authored and
charlifu
committed
[V0 Deprecation] Remove `MultiModalPlaceholderMap` (#25366)

DarkLight1337
authored and
charlifu
committed
Enable Eagle3 speculative decoding for GPT-OSS model (#25246)

eldarkurtic
authored and
charlifu
committed
[TPU][Bugfix][CI] Fix broken tests/build dependency (#25255)

NickLucche
authored and
charlifu
committed
[TPU] Deprecate `xm.mark_step` in favor of ``torch_xla.sync` (#25254)

authored and
charlifu
committed
refactor: abstract graph mode support into platform interface (#25161)

yiz-liu
authored and
charlifu
committed
[Misc] Remove unused encoder-decoder error strings (#25374)

DarkLight1337
authored and
charlifu
committed
Make pickle import check fast (#25379)

hmellor
authored and
charlifu
committed
Make `mypy` behave like a proper pre-commit hook (#25313)

authored and
charlifu
committed
[Kernel] MI-300X triton moe configs (#23445)

Sara-KS
authored and
charlifu
committed
[Bugfix] Fix several issues with p2p xPyD in GET type (#23993)

authored and
charlifu
committed
[V1][Attention] Split triton_attn in triton-only and rocm specific backends (#24648)

bringlein
authored and
charlifu
committed
[EPLB] Reduce EPLB Inference Overhead (#24573)

authored and
charlifu
committed
[CLI env var] Add VLLM_FLASH_ATTN_MAX_NUM_SPLITS_FOR_CUDA_GRAPH in env variables (#25274)

authored and
charlifu
committed
[Compiler] Disable Inductor standalone compile by default (#25391)

ElizaWszola
authored and
charlifu
committed
[CI Failure] Fix fp8 kv cache on <SM90 (#25396)

mgoin
authored and
charlifu
committed
[DP] support torchrun external launcher with Data Parallelism (#24899)

authored and
charlifu
committed
[misc] Remove RFC review hours reference (#25416)

simon-mo
authored and
charlifu
committed
[torch.compile] Cleanup compilation tests and custom passes, add debug utils, fix DCE bug (#23091), fix test (#24376), and prep for custom op matching (#24604) (#24542)

ProExpertProg
authored and
charlifu
committed
[KV offload][5/N] Add `CPUOffloadingSpec` (#24251)

orozery
authored and
charlifu
committed
[CI/Build] Skip Qwen3-VL initialization tests until models are actually released (#25394)

DarkLight1337
authored and
charlifu
committed
[TPU] update torch_xla dependency for PyPI compatibility (#25278)

authored and
charlifu
committed
[Frontend] Responses API MCP tools for built in tools and to pass through headers (#24628)

authored and
charlifu
committed
[Bugfix] fix custom op test (#25429)

ProExpertProg
authored and
charlifu
committed
[Core] Drop overly aggressive whisper assertion (#25408)

russellb
authored and
charlifu
committed
[Bugfix] Fix missing `clear_connector_metadata` (#25397)

NickLucche
authored and
charlifu
committed
[BugFix] [DP/EP] Fix slow execution when BS <= DP (#25407)

authored and
charlifu
committed
[Performance] Remove input pads in cutlass_mla and optimize v_proj output handling (#25184)

alexm-redhat
authored and
charlifu
committed
[Perf] Apply torch.compile for `per_block_cast_to_fp8` (#24611)

yewentao256
authored and
charlifu
committed
[V0 deprecation] Remove platform v1 controling interface (#25410)

Isotr0py
authored and
charlifu
committed
[V0 deprecation] Remove `_set_default_args_v0` function (#25409)

Isotr0py
authored and
charlifu
committed
[Bug] Fix Long Context OOM Issue (#25290)

yewentao256
authored and
charlifu
committed
[feat] Support MRoPE + YaRN (#25384)

authored and
charlifu
committed
[XPU] Fix `compile_size` is `None` case. (#25433)

jikunshang
authored and
charlifu
committed
[benchmarks]allow skip ready check for bench serve (#25420)

authored and
charlifu
committed
[Bugfix] Remove contiguous output req for context parallel MLA (#25414)

mgoin
authored and
charlifu
committed
[Docs] Fix griffe warnings in vllm/lora/ops (#25369)

windsonsea
authored and
charlifu
committed
[DP/EP][GPTOSS] Use triton matmul-ogs kernels for GPTOSS DP/EP (#24588)

authored and
charlifu
committed
[NIXL][OOT platform] support nixl_connector with oot platform and other nixl_backend (#25121)

xuechendi
authored and
charlifu
committed
[Model] Enable DP for ViT in Qwen2-VL (#25445)

DarkLight1337
authored and
charlifu
committed
Handle triton kernel import exception (#25319)

minosfuture
authored and
charlifu
committed
[Frontend] Add a new xml-based tool parser for qwen3-coder (#25028)

Zhikaiiii
authored and
charlifu
committed
[Misc] Move DP for ViT code inside model executor dir (#25459)

DarkLight1337
authored and
charlifu
committed
[Test]: Hermes tool parser stream output error in Qwen3 case (#25203)

ahartel
authored and
charlifu
committed
[Bugfix] Fix idefics3 `tie_word_embeddings` (#25454)

Isotr0py
authored and
charlifu
committed
[Core] Optimize LoRA weight loading (#25403)

jeejeelee
authored and
charlifu
committed
[docs] Benchmark Serving Incorrect Arg (#25474)

vllmellm
authored and
charlifu
committed
[CI/Build] Fix disabled v1 attention backend selection test (#25471)

Isotr0py
authored and
charlifu
committed
[BugFix] Register expert_map as named buffer for wake_up and sleep (#25458)

authored and
charlifu
committed
[P/D] Support NIXL connector to disconnect during a clean shutdown (#24423)

authored and
charlifu
committed
[Docs] NixlConnector quickstart guide (#24249)

authored and
charlifu
committed
[XPU] Fix MOE DP accuracy issue on XPU (#25465)

faaany
authored and
charlifu
committed
[UX] Change kv-cache-memory log level to debug (#25479)

mgoin
authored and
charlifu
committed
[V1] Remove V0 code paths for Hybrid models (#25400)

tdoublep
authored and
charlifu
committed
Use fusion pass to select AITER group quant RMSNorm and w8a8 gemm (#707)

micah-wil
authored and
charlifu
committed
add is rocm aiter linear enabled
charlifu
committed

Uh oh!

New PR number #25693#25688charlifu wants to merge 275 commits intovllm-project:mainvllm-project/vllm:mainfrom ROCm:amd/aiter_fusion_passROCm/vllm:amd/aiter_fusion_passCopy head branch name to clipboard

Commits

Commits on Sep 25, 2025

New PR number #25693#25688
charlifu wants to merge 275 commits intovllm-project:mainfrom
ROCm:amd/aiter_fusion_pass