Michael Goin aab549870d Use Blackwell FlashInfer MXFP4 MoE by default if available (#23008) ...

Signed-off-by: mgoin <mgoin64@gmail.com>

2025-08-18 15:27:58 -07:00

..

adapter_commons

…

assets

[Bugfix] Fix ndarray video color from VideoAsset (#21064)

2025-07-19 02:17:16 -07:00

attention

[Bugfix gpt-oss] Fix float32 convert for flashinfer sink support (#23016)

2025-08-16 11:16:00 -07:00

benchmarks

Add docs for PrefixRepetitionDataset + enable usage with vllm bench throughput (#23012)

2025-08-16 10:21:20 +00:00

compilation

[Core] Allow full cudagraph with separate attention routines and orthogonal to compilation, add support for FA2 and FlashInfer (#20059)

2025-08-15 10:01:39 -04:00

config

[V1] Logits processors extensibility (#19912)

2025-08-16 12:59:17 -07:00

core

[V0 Deprecation] Remove multi-step scheduling (#22138)

2025-08-12 20:18:39 -07:00

device_allocator

…

distributed

[Kernels] Clean up FusedMoeMethodBase and modular kernel setup. Remove extra arguments from modular kernel methods. (#22035)

2025-08-15 14:46:00 -04:00

engine

[V1] Logits processors extensibility (#19912)

2025-08-16 12:59:17 -07:00

entrypoints

fix: OpenAI SDK compat (ResponseTextConfig) (#23126)

2025-08-18 15:27:51 -07:00

executor

[executor] feat: add supports_pp attr to executors (#21786)

2025-08-03 18:04:45 +08:00

inputs

[Refactor] Defer tensor data construction in MultiModalKwargs (#23030)

2025-08-16 21:05:50 -07:00

logging_utils

…

lora

[Bugfix] Fix: Fix multi loras with tp >=2 and LRU cache (#20873)

2025-07-31 19:48:13 -07:00

model_executor

Use Blackwell FlashInfer MXFP4 MoE by default if available (#23008)

2025-08-18 15:27:58 -07:00

multimodal

[Refactor] Defer tensor data construction in MultiModalKwargs (#23030)

2025-08-16 21:05:50 -07:00

platforms

[XPU]avoid circular import during XPU init (#23017)

2025-08-16 05:16:34 +00:00

plugins

[XPU] upgrade torch 2.8 on for XPU (#22300)

2025-08-08 17:03:45 -07:00

profiler

…

ray

[Misc] Getting and passing ray runtime_env to workers (#22040)

2025-08-01 23:54:40 -07:00

reasoning

[Frontend] Avoid list copies in serving_chat.py (#22947)

2025-08-16 02:06:30 +00:00

third_party

…

transformers_utils

[Bugfix] Replace custom Encoding class with BatchEncoding in MistralTokenizer (#22786)

2025-08-14 08:09:27 -07:00

triton_utils

[Bugfix] Add missing field to TritonLanguagePlaceholder (#20812)

2025-07-11 05:25:11 -07:00

usage

[Misc] Add logger.exception for TPU information collection failures (#20510)

2025-07-05 07:24:32 -07:00

utils

[V1] Logits processors extensibility (#19912)

2025-08-16 12:59:17 -07:00

v1

[XPU] fix xpu to set cudagraph batch sizes (#23044)

2025-08-17 21:45:42 +00:00

vllm_flash_attn

…

worker

[V0 Deprecation] Remove advance_step (#22969)

2025-08-15 08:22:31 -07:00

__init__.py

[Perf][CLI] Improve overall startup time (#19941)

2025-06-22 23:11:22 +00:00

_custom_ops.py

[Kernel/Quant] Remove AQLM (#22943)

2025-08-16 19:38:21 +00:00

_ipex_ops.py

[XPU]Fix flash_attn_varlen_func interface on xpu (#22350)

2025-08-06 19:28:11 -07:00

beam_search.py

…

collect_env.py

[Misc] Refactor: Improve argument handling for conda command (#20481)

2025-07-15 19:43:19 -07:00

connections.py

…

env_override.py

…

envs.py

[Bugfix] Added more env vars to hash (#22449)

2025-08-15 20:08:37 +00:00

forward_context.py

[Core] Allow full cudagraph with separate attention routines and orthogonal to compilation, add support for FA2 and FlashInfer (#20059)

2025-08-15 10:01:39 -04:00

logger.py

Optimize logger init performance by using module-level constants (#22373)

2025-08-06 20:32:19 -07:00

logits_process.py

…

outputs.py

[Model][2/N] Automatic conversion of CrossEncoding model (#19978)

2025-07-03 13:59:23 +00:00

pooling_params.py

Support token_type_ids in V1 with less code changes (#21985)

2025-08-10 22:54:59 -07:00

py.typed

…

sampling_params.py

[Docs] Fix warnings in docs build (#22588)

2025-08-10 05:49:51 -07:00

scalar_type.py

[Kernel] [Quantization] Add MXFP4 and bias support for marlin kernel (#22428)

2025-08-14 11:23:22 -07:00

scripts.py

…

sequence.py

[Refactor] Defer tensor data construction in MultiModalKwargs (#23030)

2025-08-16 21:05:50 -07:00

tasks.py

[V1] Get supported tasks from model runner instead of model config (#21585)

2025-07-25 05:36:45 -07:00

test_utils.py

[Misc] Modify the organization of GLM series (#22171)

2025-08-03 23:51:20 -07:00

tracing.py

…

version.py

…