Commits vllm-project/vllm

[Metrics] Add Prometheus counter for CUDA graph iteration mode

tlrmchlsmth committed 50 days ago

119ccde4

[bug] Fix deadlock with pause resume and collective_rpc (#37024)

hao-aaron committed 50 days ago

Verified 6accb21f

[Model Runner V2] Spec decode rejection sampler logprobs support (#37237)

TheEpicDolphin committed 51 days ago

Verified 053f3b63

[BUG] Exclude SKIP_TENSORS from get_layer_size() + new weight sync example for dpep (#37334)

hao-aaron committed 51 days ago

Verified 5f82706a

[EPLB] Simplify EPLB rearrange by only returning one map (#36267)

SageMoore committed 51 days ago

Verified c32a58cc

[Bugfix] Zero-init MLA attention output buffers to prevent NaN from CUDA graph padding (#37442)

elvircrn committed 51 days ago

Verified ef2c4f77

[XPU]Unify xpu test dependencies in dockerfile.xpu (#36477)

1643661061leo committed 51 days ago

Verified 9dade5da

[Bugfix] Expand quantization method support in perf metrics (#37231)

thillai-c committed 51 days ago

Verified 828f862a

[Bugfix] Fix KV scales inconsistency in fp8 MLA & FlashInfer kv_cache_dtype "auto" leading to gibberish (#37054)

andylolu2 committed 51 days ago

Verified 577df69b

[Model Runner V2] Spec decode rejection sampler greedy support (#37238)

TheEpicDolphin committed 51 days ago

Verified 04244fd0

[Bugfix] Remove assertion for NVFP4 scale dynamic range (#37465)

mgoin committed 51 days ago

Verified 9482b0b0

[LoRA][BugFix] Fix skipped LoRA adapters for Mistral3 (#36928)

WoosukKwon committed 51 days ago

Verified 5bc1da14

fix(worker): optimize swap_states to copy only active token prefixes (#34733)

pjo256 committed 51 days ago

Verified 00910171

[V0 Deprecation] Deprecate virtual engine (#37195)

yewentao256 committed 51 days ago

Verified 0d81a1fe

chunk parakeet into 30s clips to prevent OOMs on long audios (#36671)

netanel-haber committed 51 days ago

Verified 6ae4c8d6

[Bugfix] Fix ROCm crash in qwen3_next multi-stream events (#36795) (#37427)

JartX committed 51 days ago

Verified a913b612

Fix models which use `layer_type_validation` for Transformers v5 (#37398)

hmellor committed 51 days ago

Verified 5ce2d10e

[Bugfix] Fix incorrect use of merge_size in Qwen3-VL video timestamp calculation (#37439)

cnyvfang committed 51 days ago

Verified 738d0a28

[bugfix][async scheduling] fix extra cuda context in device 0 with EP/DP (#37449)

youkaichao committed 51 days ago

Verified 70b81c4f

[Model] Remove unnecessary processor definition for Nemotron Parse (#37456)

DarkLight1337 committed 51 days ago

Verified 7476d148

[Misc] Clean up model registry (#37457)

DarkLight1337 committed 51 days ago

Verified f3732bd9

[Perf] Add tuned triton moe config for Qwen3.5 H200, 9.9% E2E throughput improvement (#37340)

yewentao256 committed 51 days ago

Verified 0ef7f790

[kv_offload+HMA][2/N]: Support multiple KV groups in GPULoadStoreSpec (#36642)

orozery committed 51 days ago

Verified 5dd8df07

Add API docs link if the CLI arg is a config class (#37432)

hmellor committed 51 days ago

Verified 39bfb57b

Adding deterministic lora benchmarking to vLLM Bench (#36057)

RonaldBXu committed 51 days ago

Verified c9d838fc

[Kernel] Add gpt-oss Router GEMM kernel (#37205)

xyang16 committed 51 days ago

Verified b1169d7b

standardize load_weights using AutoWeightsLoader for kimi_linear and minimax_text_01 (#37371)

XLiu-2000 committed 51 days ago

Verified 17808394

[Perf] Eliminate padding and slicing op for GPT-OSS with Flashinfer MXFP4 MXFP8 MoE (#30647)

elvischenv committed 51 days ago

Verified 296839a1

[Log] Reduce duplicate log (#37313)

yewentao256 committed 51 days ago

Verified c373b5c0

elastic_ep: Fix stateless group port races (#36330)

itayalroy committed 51 days ago

Verified de1a86b7