vLLM/vllm at d9dac2b3d4cdd57ff2f7c5311bc8aaf3eef3feec - vLLM

assets

[Bugfix] Align OpenCV video metadata timeline (#47099 )

2026-06-30 20:43:42 +00:00

benchmarks

[Benchmark] Add probe requests to vllm bench serve (#49611 )

2026-08-01 22:10:56 +08:00

compilation

[compile] Fix fake kernel return dtype (#50444 )

2026-07-30 19:50:02 +00:00

config

[Attention][MiniMax-M3] Add MSA speculative decode verification (#50032 )

2026-08-02 13:46:55 -07:00

cute_utils

Revert "[Perf][GLM-5.2] Blackwell decode optimizations" (#49768 )

2026-07-24 16:04:59 -07:00

device_allocator

[Bugfix][CuMem] Make KV-cache wake cleanup tag-safe (#49857 )

2026-07-26 12:09:13 -07:00

distributed

[Bugfix][KV Connector] Propagate EAGLE state across merged Mooncake store groups (#49069 )

2026-08-02 19:32:26 -07:00

engine

[Multimodal] Expose mm hash algothrim selection to cli args (#49686 )

2026-07-31 05:25:19 +00:00

entrypoints

[Frontend] Require cache_salt to be non-empty via schema (#50816 )

2026-08-03 06:51:27 +00:00

inputs

[Frontend][Core] Standardize request error handling with VLLMError hierarchy (#49665 )

2026-07-29 04:20:40 +00:00

ir

[CI] Use explicit devices in IR tests (#49513 )

2026-07-23 20:08:14 -07:00

kernels

[XPU] Unify XPU RMSNorm kernels with vllm_c and drop redundant XPU-specific implementation (#46981 )

2026-07-31 17:28:18 +08:00

logging_utils

[vLLM IR] 1/N Implement IR skeleton and rms_norm op (#33825 )

2026-03-31 22:15:05 -04:00

lora

[MoE Refactor] Rename FusedMoE to FusedMoEFactory (#44941 )

2026-07-31 11:28:18 +01:00

model_executor

Shard the K3 Latent-MoE up-projection on large batches (#50383 )

2026-08-03 00:49:39 -07:00

models

Shard the K3 Latent-MoE up-projection on large batches (#50383 )

2026-08-03 00:49:39 -07:00

multimodal

[Perf] Speed up multimodal placeholder and token-match scanning (#50716 )

2026-08-03 08:27:56 +00:00

parser

[GPT-OSS] Strict tool call and constrained decoding for Harmony (#45560 )

2026-08-01 12:10:26 -04:00

platforms

[ROCm][ViT] Detect Triton-AMD kernels at their new aiter location (#40289 )

2026-07-31 15:30:09 -04:00

plugins

[Frontend] Add endpoint plugins framework (#47454 )

2026-07-08 10:00:41 +08:00

profiler

[Refactor] Clean up log once scope="local" (#40540 )

2026-04-22 16:42:43 -04:00

ray

Propagate Flash Attention cache configuration to Ray workers (#49177 )

2026-07-21 07:47:53 +00:00

reasoning

[GPT-OSS] Strict tool call and constrained decoding for Harmony (#45560 )

2026-08-01 12:10:26 -04:00

renderers

[GPT-OSS] Strict tool call and constrained decoding for Harmony (#45560 )

2026-08-01 12:10:26 -04:00

third_party

[New model] Kimi K3 (#50000 )

2026-07-30 03:49:58 -07:00

tokenizers

[Frontend] Cohere chat v2 api support (#47189 )

2026-07-31 22:53:54 -07:00

tool_parsers

[GPT-OSS] Strict tool call and constrained decoding for Harmony (#45560 )

2026-08-01 12:10:26 -04:00

tracing

…

transformers_utils

[New model] Kimi K3 (#50000 )

2026-07-30 03:49:58 -07:00

triton_utils

[ROCm] [Model] Enable TML inkling (#48841 )

2026-07-27 10:05:46 +00:00

usage

replace cuda_device_count_stateless() to current_platform.device_count() (#37841 )

2026-03-31 22:32:54 +08:00

utils

[UX] Reduce startup log noise (#50590 )

2026-07-31 13:04:34 -07:00

v1

cpu_model_runner.py: skip the warm up if CompilationMode.NONE (#50547 )

2026-08-03 04:34:29 +00:00

vllm_flash_attn

[2/N][Attention] Enable masked MHA for sparse MLA prefills (#48770 )

2026-07-31 06:01:24 -07:00

__init__.py

…

_aiter_ops.py

Enable gfx1250 ROCm architecture (#46516 )

2026-07-31 01:41:12 +00:00

_custom_ops.py

[Attention][MiniMax-M3] Add MSA speculative decode verification (#50032 )

2026-08-02 13:46:55 -07:00

_xpu_ops.py

[XPU][DeepSeekV4]Add DeepSeek-V4 fuse_index_q SYCL kernel path (#45991 )

2026-07-21 21:22:34 +08:00

collect_env.py

[XPU]Enhance environment collection for Intel XPU and optimize layout (#35698 )

2026-04-13 12:51:46 +08:00

connections.py

[CI] Add retry with 4x backoff to HTTP fetches for transient failures (#37218 )

2026-03-19 19:00:20 +00:00

env_override.py

[Perf] Add VLLM_TRITON_FORCE_FIRST_CONFIG to skip Triton autotuning (#42425 )

2026-06-16 15:16:45 +02:00

envs.py

[Frontend] Cohere chat v2 api support (#47189 )

2026-07-31 22:53:54 -07:00

exceptions.py

[Frontend][Core] Standardize request error handling with VLLMError hierarchy (#49665 )

2026-07-29 04:20:40 +00:00

forward_context.py

[Feature] Support sequence parallel without the need for DP, 1.9%~5.0% E2E Throughput Improvement (#47070 )

2026-07-05 05:41:30 -07:00

logger.py

[Log] Log once in local node by default (#37568 )

2026-03-19 12:04:59 -07:00

logits_process.py

…

logprobs.py

…

model_inspection.py

…

outputs.py

[Front-end] [Messages] Populate num_cache_creation_tokens (#48535 )

2026-07-18 13:04:35 -04:00

pooling_params.py

[Frontend][Core] Standardize request error handling with VLLMError hierarchy (#49665 )

2026-07-29 04:20:40 +00:00

py.typed

…

sampling_params.py

[Frontend][Core] Standardize request error handling with VLLMError hierarchy (#49665 )

2026-07-29 04:20:40 +00:00

scalar_type.py

Humming support for 2/3/5/6/7-bit pack-quantized weight-only inference (#46389 )

2026-06-24 13:53:54 -04:00

scripts.py

…

sequence.py

[Core] Cleanup KVConnector handling with PP + fix MRV2 (#43732 )

2026-05-28 13:12:03 -07:00

tasks.py

[Bugfix] Reject removed pooling parameters (#48984 )

2026-07-19 05:18:03 -07:00

version.py

…