[Misc] Add backup hash algorithm for FIPS constrained environments#28477

Closed

geodavic wants to merge 180 commits intovllm-project:mainfrom

Commits on Nov 16, 2025

use safe hash
geodavic
committed
better exception handling
geodavic
committed
lint
geodavic
committed
[Kernel] Optimize rms_norm kernel (#27931)

xyang16
authored and
geodavic
committed
[BugFix] Fix Siglip2Attention on XPU (#28448)

faaany
authored and
geodavic
committed
[Misc] Remove unused attention prefix prefill ops functions (#26971)

lgeiger
authored and
geodavic
committed
[Perf] Use np.ndarray instead of list[list[int]] to reduce GC overhead (#28245)

Jialin
authored and
geodavic
committed
[V0 deprecation] Clean up num_prefill_tokens logic for V0 (#28203)

gcanlin
authored and
geodavic
committed
[Misc] fix typo in DCP comment (#28389)

Livinfly
authored and
geodavic
committed
[LoRA][1/N]Remove LoRA extra vocab (#28382)

jeejeelee
authored and
geodavic
committed
[TPU] Rename path to tpu platform (#28452)

kyuyeunk
authored and
geodavic
committed
[Misc] Cleanup Executor interface (#28441)

wangxiyuan
authored and
geodavic
committed
Add Zurich vLLM Meetup (#28488)

mgoin
authored and
geodavic
committed
[Bugfix] Disable shared expert overlap if Marlin MoE is used (#28410)

mgoin
authored and
geodavic
committed
[Feature] Allow configuring FlashInfer workspace size (#28269)

authored and
geodavic
committed
Use FLASHINFER MLA backend when testing fp8_kv_scale_compile (#28491)

adabeyta
authored and
geodavic
committed
[BugFix] Graceful handling of torch symm mem errors. (#27671)

authored and
geodavic
committed
[Frontend] Change CompilationMode to a proper Enum (#28165)

gmagogsfm
authored and
geodavic
committed
[Performance] Cache loaded custom logitsprocs to avoid overheads (#28462)

Isotr0py
authored and
geodavic
committed
[[V0 deprecation]]Remove VLLM_USE_V1 env (#28204)

wangxiyuan
authored and
geodavic
committed
[CPU] Refactor CPU attention backend (#27954)

bigPYJ1151
authored and
geodavic
committed
`VLLM_USE_TRITON_FLASH_ATTN` V0 variable deprecation (#27611)

AndreasKaratzas
authored and
geodavic
committed
[Model][Qwen3VL] Simplify `get_mrope_input_positions` using numpy (#28302)

lgeiger
authored and
geodavic
committed
[Core] Encoder separation for Encode-Prefill-Decode Disaggregation (#25233)

authored and
geodavic
committed
[BugFix] Add fallback path in `apply_rotary_pos_emb_flashattn` for non-cuda platforms (#28447)

faaany
authored and
geodavic
committed
[Benchmark] Add retry support to fix workload bias in multi-turn benchmark (#28493)

ai-jz
authored and
geodavic
committed
[Core] Cache `vllm_is_batch_invariant` (#28304)

lgeiger
authored and
geodavic
committed
[CI/Build] Fix crash due to removed VLLM_USE_V1 attribute in EPD (#28521)

authored and
geodavic
committed
[CI] Introduce autorun_on_main feature (#27836)

hl475
authored and
geodavic
committed
[BugFix]: --enable-lora with model granite-4.0-micro crash (#27733)

yyzxw
authored and
geodavic
committed
[Model] fix glm4_moe_mtp load weights with GLM-4.6 checkpoint. (#27597)

authored and
geodavic
committed
[XPU]Fix crash due to removed VLLM_USE_V1 attribute (#28520)

chaojun-zhang
authored and
geodavic
committed
[KVConnector] Enable get_block_ids_with_load_errors() in LMCache connector (#27978)

authored and
geodavic
committed
add cpu option for p/d in nixl_connector (#28356)

ZhengHongming888
authored and
geodavic
committed
[ROCm] [Bugfix] Fix `fused_qknorm_rope_kernel` rocm compatibility (#28500)

tjtanaa
authored and
geodavic
committed
[Bugfix] Fix gpt_oss packed_modules_mapping (#28536)

jeejeelee
authored and
geodavic
committed
[V0 deprecation] Deprecate use_v1 parameter (#28112)

wangxiyuan
authored and
geodavic
committed
Fix pre-commit (and XPU) on `main` (#28556)

hmellor
authored and
geodavic
committed
[Performance][Hopper] Avoid M dim padding to 4x for most cases (due to cuda graphs paddings) (#28492)

alexm-redhat
authored and
geodavic
committed
[Refactor] Remove redundant TP gather/split in split_qkv in QwenVL (#28271)

gcanlin
authored and
geodavic
committed
[Misc] Refactor Attention kv transfer methods into decorator (#27816)

authored and
geodavic
committed
Remove deprecated fields from `CompilationConfig` (#27593)

hmellor
authored and
geodavic
committed
[Perf] Refactor cudagraph_support to enable full CUDA graphs for spec decoding with FlashInfer (#28479)

benchislett
authored and
geodavic
committed
Implement ARC KV cache eviction policy for CPU offloader (#27039)

authored and
geodavic
committed
[EPLB][ROCm]: support EPBL for ROCm backend (#27731)

authored and
geodavic
committed
[Model] [Config] Correctly identify granite-4.0-micro as non-hybrid model (#28563)

tdoublep
authored and
geodavic
committed
[CI] Skip "Multi-Modal Models Test (Extended) 3" test that's broken in current Transformers (#28559)

hmellor
authored and
geodavic
committed
[KV connector][WIP] KV cache proxy based on LMCache multi-process mode (#27902)

ApostaC
authored and
geodavic
committed
[BugFix] Priority scheduling and spec tokens preemption (#28558)

andylolu2
authored and
geodavic
committed
[Misc]Fix typo in llm_engine.py (#28584)

frank-wei
authored and
geodavic
committed
[Performance][B200] Fix deepgemm prologue (#27897)

authored and
geodavic
committed
[ROCM] Fix ROCm warnings, environment flag access, and GEMM kernel naming for consistency in `_aiter_ops.py` (#28464)

vllmellm
authored and
geodavic
committed
[TPU] Support GCS path in VLLM_TORCH_PROFILER_DIR (#28487)

QiliangCui
authored and
geodavic
committed
[Bugfix] Adjust Marlin CUDA arch selection to 8.0+PTX;9.0+PTX (#28294)

mgoin
authored and
geodavic
committed
[Core][AMD] Migrate fully transparent sleep mode to ROCm platform (#12695)

authored and
geodavic
committed
[MoE][Kernel][Perf] Improve Shared Expert Stream Overlap (#28406)

alexm-redhat
authored and
geodavic
committed
Skip models that cannot currently init on Transformers v5 (#28471)

hmellor
authored and
geodavic
committed
[Docs] Update meetups.md description (#28583)

mgoin
authored and
geodavic
committed
[ROCm][Bugfix] Revert removing setuptools version restriction (#28592)

gshtras
authored and
geodavic
committed
[platform] Move get_cu_count to utils (#27005)

wangxiyuan
authored and
geodavic
committed
[Bugfix] Fix SM100 gpt-oss regression due to faulty attn sink support (#28561)

mgoin
authored and
geodavic
committed
[BugFix] Fix `mm_encoder_attn_backend` arg type checking (#28599)

njhill
authored and
geodavic
committed
[Docs] Add some details about what the MoE block needs for the Transformers backend (#28588)

hmellor
authored and
geodavic
committed
Rename clashing method names for vLLM model protocol (#27583)

hmellor
authored and
geodavic
committed
[n-gen] DO NOT repeatedly return finished child requests (#28591)

Jialin
authored and
geodavic
committed
[Frontend] split append tool output (#28333)

authored and
geodavic
committed
[Frontend][responsesAPI][1/n] convert responses API tool input to chat completions tool format (#28231)

authored and
geodavic
committed
[BugFix][ROCm] Fix `get_cu_count` missing variable error (#28608)

ganyi1996ppo
authored and
geodavic
committed
[XPU] Support Triton path for LoRA operations on XPU (#28511)

faaany
authored and
geodavic
committed
Support DeepEP for Kimi-k2-thinking through enabling gemm selection for compressed-tensor marlin wna16 (#28574)

luccafong
authored and
geodavic
committed
[build][cmake]: Bundle static ACL and torch libgomp for CPU extension builds (#28059)

Radu2k
authored and
geodavic
committed
[ROCm][BugFix] Remove the usage of `device_info` from aiter (#28383)

ganyi1996ppo
authored and
geodavic
committed
[Bugfix] Prevent crash on empty grammar string (#28210)

tjandy98
authored and
geodavic
committed
Use official xformers-0.0.33 built for PT 2.9 (#28600)

authored and
geodavic
committed
Add NUMA node validation for CPU thread binding (#28555)

usberkeley
authored and
geodavic
committed
[Bugfix] fix kimi-linear crash (#28445)

ZJY0516
authored and
geodavic
committed
[Frontend] supports interleaved thinking (#28531)

chaunceyjiang
authored and
geodavic
committed
Support all interleaved layer types (#28485)

sarckk
authored and
geodavic
committed
Fix: Correctly filter special tokens in benchmark_prefix_caching (#28615)

dw2761
authored and
geodavic
committed
[BugFix] Fix type error when assign a trition kernel tensor to a torch.nn.Parameter (#28603)

liuzijing2014
authored and
geodavic
committed
Fix io processor pooling #28273 (#28484)

baonudesifeizhai
authored and
geodavic
committed
[XPU] add sym params to IPEXConfig (#28611)

zufangzhu
authored and
geodavic
committed
[Bugfix] Fix FPS value type for Qwen2.5-Omni video processing (#28630)

faaany
authored and
geodavic
committed
[Hardware][PowerPC] Fix fp16 compilation error for Power in cpu attention backend and bump oneDNN version (#28535)

authored and
geodavic
committed
[ROCm][BugFix]Fix `get_cu_count` in rocm_aiter_fa.py (#28618)

ganyi1996ppo
authored and
geodavic
committed
[CI/Build] Install uv for AMD MI300: Language Models Tests (Hybrid) %N (#28142)

authored and
geodavic
committed
[CI Failure] Fix backend selection for encoder-only models (#28534)

hl475
authored and
geodavic
committed
[BugFix] DeepSeek-OCR: apply NoRepeatNGramLogitsProcessor to greedy path (#28617)

YuanpingSong
authored and
geodavic
committed
Fix `get_num_experts` when config sets it explicitly to `None` (#28652)

authored and
geodavic
committed
[Misc] Turn off encoder torch compile by default (#28634)

ywang96
authored and
geodavic
committed
Rewrite C++ meta funcs to Python (#28595)

janeyx99
authored and
geodavic
committed
[BugFix] Ensure `EngineArgs.create_engine_config` is idempotent (#28515)

njhill
authored and
geodavic
committed
[TPU] patch TPU wheel build script to resolve metadata issue (#27279)

jcyang43
authored and
geodavic
committed
[Performance][B200] silu_mul_quant: pack scales in int32 (#28358)

authored and
geodavic
committed
[Bugfix] Fix validate model input for decoder models (#27099)

authored and
geodavic
committed
[Attention][Bugfix] Fix FA sink support (#28660)

MatthewBonanni
authored and
geodavic
committed
[Perf] Support stream interval for reducing host overhead (#27869)

authored and
geodavic
committed
[bugfix] correct local_chunk_len for DCP in reorg_kvcache with long context (#28526)

authored and
geodavic
committed
[Bugfix] Eliminate tuple inputs to submodules in graph partitioning (#28533)

gmagogsfm
authored and
geodavic
committed
[Bugfix] [CPU] bump torch to 2.9.0 for Darwin to fix segmentation fault (#27791)

authored and
geodavic
committed
[Misc] Update CODEOWNERS for simon-mo and comaniac (#28675)

simon-mo
authored and
geodavic
committed
[CI] Bug: Fix ci entrypoint pooling (#28684)

yewentao256
authored and
geodavic
committed
[KV Connector] Test async mode in scheduler tests (#28550)

markmc
authored and
geodavic
committed
Mirrored test group definitions for AMD (2025-11-11) (#28573)

Alexei-V-Ivanov-AMD
authored and
geodavic
committed
[quantization][config] enable override existing quant_config (#28510)

authored and
geodavic
committed
[ROCm] Bump up the version of amd-smi to 6.4.3 (#28680)

SageMoore
authored and
geodavic
committed
[CPU][Bugfix] Fix Apple Silicon M1 compilation failure (#28681)

mgoin
authored and
geodavic
committed
[ci][amd] fix basic models extra init test (#28676)

bradleyhd
authored and
geodavic
committed
[Misc] Remove `warn_for_unimplemented_methods` (#28613)

DarkLight1337
authored and
geodavic
committed
[XPU][CI]disable lm cache uts (#28696)

jikunshang
authored and
geodavic
committed
[Misc] Update xformers to 0.33.0.post1 (#28678)

ywang96
authored and
geodavic
committed
[Misc] add ignore mapper for quark quantization (#28275)

haoyangli0109
authored and
geodavic
committed
[Bugfix][CI/Test][Spec Decode] Fix illegal memory access in offline_inference/spec_decode.py (Issue 27619) (#28432)

authored and
geodavic
committed
[BugFix][CI/Build][ROCM] Fix import error and apply assert in appropriate case in test_struct_output_generate (#28311)

authored and
geodavic
committed
use default CCL_ZE_IPC_EXCHANGE (#28700)

yma11
authored and
geodavic
committed
[Bugfix] fix dots.ocr pp support (#28705)

ZJY0516
authored and
geodavic
committed
[BugFix] Fix multi-modal async scheduling race condition (#28706)

njhill
authored and
geodavic
committed
Add output token counting to gsm8k eval (#28594)

mgoin
authored and
geodavic
committed
[Minor] avoid register new custom and just import silly_attn (#28578)

BoyuanFeng
authored and
geodavic
committed
[Misc] fix comment in test_envs (#28529)

xingliu14
authored and
geodavic
committed
[Metrics] Log number of preempted requests (#28522)

610lyn
authored and
geodavic
committed
[Frontend] Added chat-style multimodal support to /classify. (#27516)

authored and
geodavic
committed
[Model][MM] Extract conv layer as CustomOp (#28455)

authored and
geodavic
committed
[DCP] Support Decode Context Parallel (DCP) for GQA with Flashinfer (#25438)

authored and
geodavic
committed
Fix KV sharing fast prefill with cudagraph enabled (#28537)

authored and
geodavic
committed
[BugFix] Fix FA3 IMA with FULL_AND_PIECEWISE and cascade attention (default) (#28702)

LucasWilkinson
authored and
geodavic
committed
[Doc] Fix macOS installation dependency resolution issue (#26721)

shahfasal
authored and
geodavic
committed
[Model] Fix bailing_moe accuracy problem (#28277)

zhaozx-cn
authored and
geodavic
committed
[Bugfix][Nixl] Fix kernel physical<>logical block_size issue (#28677)

NickLucche
authored and
geodavic
committed
[Config] Clean up SchedulerConfig initialization (#28665)

DarkLight1337
authored and
geodavic
committed
[Kernels] Enable FlashInfer FP8 Blockscale on SM90 (for TEP DSR1) (#27134)

authored and
geodavic
committed
[Fix] improve aspect ratio in dummy image generation and add common VLM tests for PaddleOCR-VL (#28711)

dongbo910220
authored and
geodavic
committed
[Docs] Update the name of `Transformers backend` -> `Transformers modeling backend` (#28725)

hmellor
authored and
geodavic
committed
[CI][CPU] Smoke test for Apple Silicon using GHA MacOS runner (#28688)

mgoin
authored and
geodavic
committed
[DisaggEverything] Tokens in<>out `/generate` endpoint (#24261)

authored and
geodavic
committed
[Attention] Bump FA for removed method (#28429)

authored and
geodavic
committed
Fix typo in comment: existance -> existence (#28737)

OthmanMohammad
authored and
geodavic
committed
Remove audio optional dependency for mistral-common (#28722)

authored and
geodavic
committed
[kernel] Improve FP8 PTPC on Hopper for larger shapes (#28692)

czhu-cohere
authored and
geodavic
committed
docs(lora_resolvers): clarify multi-resolver order and storage path requirement (#28153)

authored and
geodavic
committed
LLaMA4 LoRA Adapter Enablement (#28602)

authored and
geodavic
committed
[Bugfix] [ROCm] [AITER]: Fix aiter block quant not compatible with torch compile dynamo (#28716)

tjtanaa
authored and
geodavic
committed
[Docs] Enable some more markdown lint rules for the docs (#28731)

hmellor
authored and
geodavic
committed
[Chore] Rename `SchedulerConfig.chunked_prefill_enabled` (#28735)

DarkLight1337
authored and
geodavic
committed
[Bugfix] resolve Qwen3-VL GPTQModel quantized model loading failure (#28663)

authored and
geodavic
committed
[BugFix] Fix misprint introduced by modular_kernel refactoring. (#28728)

halyavin
authored and
geodavic
committed
[ROCm][Bugfix] Fix compilation errors with fused_qknorm_rope_kernel.cu (#28682)

SageMoore
authored and
geodavic
committed
[CI] Fix macos smoke test uv cache issue (#28736)

mgoin
authored and
geodavic
committed
[Bugfix] TypeError: 'NoneType' object is not callable (#27410)

mostrowskix
authored and
geodavic
committed
[ROCm][CI/Build] Change install location of uv (#28741)

gshtras
authored and
geodavic
committed
Avoid bytecode hook and simplify TorchCompileWrapperWithCustomDipatch (#25110)

laithsakka
authored and
geodavic
committed
[Bugfix] Fix incorrect use of hidden_states for shared_experts due to do_naive_dispatch_combine (#28740)

alexm-redhat
authored and
geodavic
committed
[Bugfix] Fix ChunkedLocalAttention CUDA Graph setting (#28739)

benchislett
authored and
geodavic
committed
[Hybrid] [Kernel] Fix chunk scan kernel when BLOCK_SIZE_DSTATE > 128 (#28295)

tdoublep
authored and
geodavic
committed
[Log] Save profiler results to file instead of stdout (#28144)

authored and
geodavic
committed
[ROCm][CI/Build] Upgrade to ROCm 7.1 and AITER main (#28753)

gshtras
authored and
geodavic
committed
[Test] Rework e2e async scheduling tests (#28744)

njhill
authored and
geodavic
committed
[Core] Performance: Use list[np.ndarray] instead of list[list[int]] for output tokens for GC optimization (#26368)

Jialin
authored and
geodavic
committed
[TPU] Fix import error in tpu launch (#28758)

QiliangCui
authored and
geodavic
committed
[Model][Qwen3VL] Use `mm_position` to compute mrope positions (#28730)

authored and
geodavic
committed
[Bugfix] Build hadacore kernels on >SM90 (#28748)

mgoin
authored and
geodavic
committed
Revert "[Core] Performance: Use list[np.ndarray] instead of list[list… (#28773)

njhill
authored and
geodavic
committed
Fix IntermediateTensors initialization and add type hints (#28743)

OthmanMohammad
authored and
geodavic
committed
[NIXL] heterogeneous block_size support (#26759)

authored and
geodavic
committed
[Performance][DeepGEMM] Estimate expected_m (#28694)

authored and
geodavic
committed
[Redo] #26368 (#28771)

authored and
geodavic
committed
[RL] [V1] Remove unused device argument from reset_kv_cache (#28766)

zhuohan123
authored and
geodavic
committed
Use narrow over indexing in `hadacore_transform` to prep for ABI stable (#28756)

janeyx99
authored and
geodavic
committed
[Kernel][Moe Configs] llama4 maverick fp8 moe config tp8 on mi325 (#28709)

zhewenl
authored and
geodavic
committed
[Misc] Make `SchedulerConfig.max_model_len` init-only (#28733)

DarkLight1337
authored and
geodavic
committed
[PERF] Remove TRTLLM Gen attn kernel limitation `max_seq_len <=131072` (#28755)

vadiklyutiy
authored and
geodavic
committed
[compile] Enable sequence parallelism matching w/o custom ops enabled (#27126)

authored and
geodavic
committed
Allow Gemma3 to take image embeddings (#28483)

tingtingtangmeta
authored and
geodavic
committed
[Doc] Fix failing doc build (#28772)

authored and
geodavic
committed
[Model] Fix lmhead init bug of bailing_moe (#28777)

authored and
geodavic
committed
Add support for Eagle with separate lm-head and embed_tokens layers (#28549)

eldarkurtic
authored and
geodavic
committed
[CI] Fix broken pipeline (#28781)

njhill
authored and
geodavic
committed
[Model][Qwen3VL] Cache positional embedding indices (#28475)

authored and
geodavic
committed
[Doc]: fix typos in various files (#28567)

didier-durand
authored and
geodavic
committed
[BugFix] Fix `AssertionError: DCP not support reorder_batch_threshold > 1 now.` (#28751)

LucasWilkinson
authored and
geodavic
committed