Commit Graph

Select branches

Hide Pull Requests

0.17.0take2

7snzwi-codex/change-default-logging-behavior

acc-rate

add-cuda-12.8-wheel

agent/amd-hf-offline-retry

agent/symm-mem-nvshmem

amd-ci

amd-collab-iter0

amd_dev

amd_mori

andy-neuma-ibm-smoke

as-of-2026-05-12

as-of-2026-06-02

batched_triton_fallback

bench-latency

benchmark_serving_test

bind_kv_caches

bugfix/37931-nvfp4-batched-all2all

builder-cuda-version

builder-nvcc-toolchain

bump_numba

bz/attractive-termite

bz/bridge-rust-tool-parser-to-py

bz/civil-mouse

bz/fresh-hippopotamus

bz/human-ox

bz/logprobs-newtype

bz/minimax-m3-engine-parser

bz/refactor-build-rust-setup

bz/rust-code-coverage

bz/rust-frontend-reattach

bz/sleep-metric

bz/tolerate-oov-prompt

bz/yeasty-cephalopod

chinmay-amd-snapshot

ci/build-gb10-arm64-image

ci/gitea-pr-search-compat

ci/h200-35gb-remaining-20

ci/narrow-basic-correctness-deps

ci/narrow-entrypoints-deps

ci/narrow-models-basic-deps

ci/narrow-models-language-deps

ci/narrow-models-multimodal-deps

ci/reorder-release-pipeline

claude/nervous-meitner

claude/optimize-weight-loading-7FlLd

claude/pedantic-merkle-d957af

claude/slack-session-JTjDk

claude/zen-banach

codex/37931-flashinfer-cutedsl-batched-one-sided

codex/add-pandas-and-datasets-to-requirements

codex/change-default-logging-behavior

codex/fix-cuda12-deepep-nccl

codex/fix-deepseek-v4-flashmla-cache-shape

codex/fix-flashinfer-fp8-moe-api-compat

codex/layer-parallel-plan

codex/remove-raydistributedexecutor-from-v0-engine

codex/remove-vllm-v0-engine-references-from-docs

codex/rocm-artifact-tensorizer

codex/test-ci-retry-failure

codex/triton-warmup-compile-keys

codex/update-arch-overview-md-with-vllm-v1-details

compile-only-pr1

copilot/add-sp-min-token-to-e2e-tests

copilot/fix-31e676e9-a4af-4ed2-b74d-19d27f0a57b2

copilot/fix-56244f30-e76a-41ed-beaf-3bc9de22a2c9

copilot/fix-584be906-f283-4e17-8776-c14111357ee7

copilot/fix-870996da-9146-438e-9a52-cdc6c1743086

copilot/fix-c6914add-1b66-46d0-9948-c2e7b6f2259f

copilot/fix-cudagraph-flag-combination

copilot/update-test-conftest-to-use-moe-backend

coverage-test-cosmos3

coverage-test-jais

coverage-test-kv-triton

coverage-test-tilelang

cuda-arch-fixup

cursor/VLLM-94-usage-stats-v2-design-584f

cursor/main-branch-failure-triage-f8d5

cursor/test-quality-improvements-eeea

cutlass_fa3_mla_sparse

dbo-cudagraph-size-cherry

deep_full_cudagraph_fix

deepep-v2-integration

deepep_tweaks

deepseek_optimizations_alex_rob

dependabot/github_actions/actions/setup-python-6.2.0

dependabot/github_actions/actions/setup-python-6.3.0

dependabot/github_actions/astral-sh/setup-uv-8.2.0

dependabot/github_actions/astral-sh/setup-uv-8.3.0

dependabot/github_actions/astral-sh/setup-uv-8.3.2

dependabot/pip/fsspec-2026.4.0

dependabot/pip/fsspec-2026.6.0

dependabot/pip/minor-update-849870308e

dependabot/pip/protobuf-7.34.1

dependabot/pip/protobuf-7.35.1

dependabot/pip/pyrate-limiter-4.1.0

dependabot/pip/quack-kernels-gte-0.4.1

deprecate-timeout

disable-image-build-per-commit

disable-sd

dockerignore_deps

dsv4-pd-fixes

dsv4-routing-from-trttllm

embed-norm-fusion

extensible-kv-cache

feat-k2.5-support

feat/tokenspeed_mla_upstream

fix-doc-build

fix-gguf-test

fix-hashing-partial-blocks

fix-mig-nvml-workaround

fix-mtp

fix-mtp-dummy-run-assertion

fix-nixl-dockerfile

fix-precommit

fix/deepseek-v4-nvidia-mhc-warmup

fix/eplb-balancedness-metric

fix/eplb-debug-logging

fix/eplb-nvfp4-modelopt

fix/eplb-prometheus-metrics

fix/flashinfer-nvfp4-cross-row-scale-corruption

fix/topk-hash-indices-dtype

fix_ds_eagle

fix_fi_cutlass

fix_moe_test_flashinfer

fix_nixl_get_finished_handshake_failure

fix_nixl_triton_attn

fix_use_ep

fp8_ep_dp

full_cudagraph

gb200-0317

gemma3n-mm

gemma4-mtp

ghsa-mcmc-2m55-j8jj

glm5--fp32-router

gpu-ids

gpu_ids2

il_tool

indexer_multistream

integrate_aiter_batched_deepgemm

jax-tpu

k3-dspark-ar-fusion

k3-release

kernel-block-size-alignment-ssm

khluu-patch-1

khluu/0190-540

khluu/2/releases/v0.16.0

khluu/automate-release-dockerhub-push

khluu/b200-k8s-ci-smoke-20260429

khluu/b200-k8s-job-fixes

khluu/b200_k8s

khluu/build0405

khluu/cherrypick37322

khluu/disable_h200_x8

khluu/feb11

khluu/gemma2

khluu/gemma3

khluu/glm5

khluu/group_commands

khluu/h200

khluu/mig

khluu/mig-small-model-swaps

khluu/release-registry-cache

khluu/release-v0.20.1-uv-python

khluu/releases/v0.16.0

khluu/rocm_gemma

khluu/test_ami

khluu/trigger-perf-eval-nightly

khluu/vllm-base-uv-python

kimi-k3

longcat-2.0

lora-test

low_latency_opt

luka/fix-rms-quant-non-contiguous

luka/vllm-ir-nits

luka/vllm-ir/compile-op

luka/vllm-ir/rms-norm-batch-invariant

luka/vllm-ir/triton

lwilkinson/kv-layout/core-standardize

main

mamba_tests

manual-act-quant-fusion-llama

maybe_fix_hang_2

mergify/houseroad/config-update

mgoin/fix-piecewise-capture-attention

migrate-gpu1-to-h200-18gb-remaining

migrate-safe-jobs-to-h200-mig

minimax-m3-perf

minus_x

mla_cuda_graphs

mla_decode_any_head

mnnvl_kv_transfer

moondream2

mrv2-ci-test

openai226

optimize-prefix-caching-scheduling

pcp-alt

pd_scheduling

perf/push-allreduce-2buffer

pil_image

pr-44891

prometheus-cudagraph-pct

qwen25vl

qwen3_5_fp8

rebase-fa3-mla-sparse

rebased_fi_moe

redhat-h100-testing

reduce_scatter_comm

release

releases/v0.10.0

releases/v0.10.1

releases/v0.10.2

releases/v0.11.0

releases/v0.11.1

releases/v0.11.2

releases/v0.12.0

releases/v0.13.0

releases/v0.14.0

releases/v0.14.1

releases/v0.15.0

releases/v0.16.0

releases/v0.17.0

releases/v0.17.1

releases/v0.18.0

releases/v0.18.1

releases/v0.19.0

releases/v0.19.1

releases/v0.20.0

releases/v0.20.1

releases/v0.20.1-python-from-source

releases/v0.20.2

releases/v0.21.0

releases/v0.22.0

releases/v0.22.1

releases/v0.23.0

releases/v0.24.0

releases/v0.25.0

releases/v0.25.1

releases/v0.26.0

releases/v0.9.0

releases/v0.9.1

releases/v0.9.2

remove-fp4-moe-env-var-clean

remove_mamba_ssm

revert-21550-chengji/fix-ci

revert-22299-main

revert-26740-wentao-optimize-startup-log-2

revert-27600-torch-utils-import

revert-40687-matthias.skinny-gemm-n5

revert-batch-kv-cache-swap-38460

rocm_silu_mul_quant

running-deque

simon-mo-patch-1

skip-lmfe-tests

sm103

speed-up-sp-tests

speed-up-sp-tests-v2

split_kv_cache_init

support_global_dp_logging

test-debug-lb

tml-inkling

tms/distributed_timeout

tms/fix-nan

tms/nvfp4-nan-contamination-test

tokenspeed

topk_id_hack

torch_dynamo

tpu_v1

tpu_v1_optimized

ucx_oneshot_ar

update_from_kv_xfer_finished_race_fix

upgrade-cutedsl

upgrade-transformers-compressed-tensors

v0.16.0-before210

v0.16.0-cu128

v0.16.0-torch291

v0.8.0

v0.8.1

v0.8.2

v0.8.3

v0.8.4

v0.8.5

v1_fix_profiler

vadim/qwen35-no-deppgemm

vllm-dashboard

wentao-cache-is_sleep

wentao-dcp-support-for-v2

wentao-deprecate-embed&token_classify

wentao-enable-flashinfer-moe-fp4-by-default

wentao-epd-support-for-MRv2

wentao-fix-amd-ci-test-others-bug

wentao-fix-ci-batch-invariant-issue

wentao-fix-ci-destroy

wentao-fix-ci-pre-commit

wentao-fix-dcp-IMA-for-v2

wentao-fix-es-v2-bug

wentao-fix-flashinfer-layout

wentao-fix-mypy-models-a-b

wentao-fix-python-install-ci-error

wentao-fix-qwen3vl-launch-bug

wentao-fix-torch-compile-issue

wentao-fix-v2-is_prefiliing

wentao-fix-v2-mrope

wentao-fix-v2-test_spec_decode_acceptance_length

wentao-fused-padding-to-kernel

wentao-model-runner-v2-support-stock-torch-compile

wentao-mrv2-enable-all-moe

wentao-optimize-async-scheduling-copy

wentao-optimize-dcp-and-add-comm-func

wentao-optimize-model-runner-v2-prepare_inputs

wentao-optimize-model-runner-v2-sampler

wentao-optimize-per-token-group-quant

wentao-optimize-pooling-by-ragged-tensor

wentao-optimize-pooling-forward

wentao-optimize-workspace-reuse

wentao-prefer-sysmem-comm

wentao-remove-dead-code

wentao-remove-multiple-dead-codes

wentao-skip-work-when-empty

wentao-support-rms-norm-uncontiguous

wentao-update-batch-invariant-docstring

wentao-update-torch-to-2.9.1

wentao-wq_b-to-cudagraph

whisper-translate

wide_ep_working_branch

wide_ep_working_branch_2

woosuk-jf

woosuk/ds-exp-2

woosuk/ds-exp-ag

woosuk/dsv4-sp

woosuk/fast-topk

woosuk/glm5-gemm

woosuk/glm5-megamoe

woosuk/gumbel-fp32-rand64

woosuk/mrv2-cudagraph-attn-fix

woosuk/mrv2-draft-gumbel-decouple

woosuk/mrv2-expert-indices

woosuk/router-nixl

woosuk/test-router

woosuk/triton-fix

worktree-agent-ae80686f16ccbc350

worktree-coverage-test-mapping

worktree-fix-ci-export-subshell

worktree-fix-cudagraph-flaky

worktree-migrate-gpu1-to-h200-18gb-mig

wye-refactor-w8a8-quant

zen5-image-build

zhuohan/extensible-kv-cache-memory

zhuohan/moe-kernel-experiment

zhuohan/redundant-pooling-check

zhuohan/remove-redundant-argument

zhuohan/remove-unnecessary-instance_id-setup

zhuohan/remove-virtual-engine

zhuohan/revert-26709

#1

#2

ci/build/22474

submission

v0.1.0

v0.1.1

v0.1.2

v0.1.3

v0.1.4

v0.1.5

v0.1.6

v0.1.7

v0.10.0

v0.10.0rc1

v0.10.0rc2

v0.10.1

v0.10.1.1

v0.10.1rc1

v0.10.2

v0.10.2rc1

v0.10.2rc2

v0.10.2rc3

v0.11.0

v0.11.0rc1

v0.11.0rc2

v0.11.0rc3

v0.11.0rc4

v0.11.0rc5

v0.11.0rc6

v0.11.1

v0.11.1rc0

v0.11.1rc1

v0.11.1rc2

v0.11.1rc3

v0.11.1rc4

v0.11.1rc5

v0.11.1rc6

v0.11.1rc7

v0.11.2

v0.12.0

v0.13.0

v0.13.0rc1

v0.13.0rc2

v0.13.0rc3

v0.13.0rc4

v0.14.0

v0.14.0rc0

v0.14.0rc1

v0.14.0rc2

v0.14.1

v0.15.0

v0.15.0rc0

v0.15.0rc1

v0.15.0rc2

v0.15.0rc3

v0.15.1

v0.15.1rc0

v0.15.1rc1

v0.15.2rc0

v0.16.0

v0.16.0rc0

v0.16.0rc1

v0.16.0rc2

v0.16.0rc3

v0.16.1rc0

v0.17.0

v0.17.0rc0

v0.17.0rc1

v0.17.1

v0.17.1rc0

v0.17.2rc0

v0.18.0

v0.18.0rc0

v0.18.0rc1

v0.18.0rc2

v0.18.1

v0.18.1rc0

v0.18.2rc0

v0.19.0

v0.19.0rc0

v0.19.0rc1

v0.19.1

v0.19.1rc0

v0.19.2rc0

v0.2.0

v0.2.1

v0.2.1.post1

v0.2.2

v0.2.3

v0.2.4

v0.2.5

v0.2.6

v0.2.7

v0.20.0

v0.20.0rc1

v0.20.1

v0.20.1rc0

v0.20.2

v0.20.2rc0

v0.21.0

v0.21.0rc1

v0.21.0rc2

v0.21.0rc3

v0.21.1rc0

v0.22.0

v0.22.0rc1

v0.22.0rc2

v0.22.0rc3

v0.22.1

v0.22.1rc0

v0.22.1rc1

v0.22.1rc2

v0.23.0

v0.23.0rc2

v0.23.1rc0

v0.24.0

v0.24.0rc1

v0.24.0rc2

v0.25.0

v0.25.0rc1

v0.25.0rc2

v0.25.0rc3

v0.25.1

v0.26.0

v0.26.0rc1

v0.26.1rc0

v0.3.0

v0.3.1

v0.3.2

v0.3.3

v0.4.0

v0.4.0.post1

v0.4.1

v0.4.2

v0.4.3

v0.5.0

v0.5.0.post1

v0.5.1

v0.5.2

v0.5.3

v0.5.3.post1

v0.5.4

v0.5.5

v0.6.0

v0.6.1

v0.6.1.post1

v0.6.1.post2

v0.6.2

v0.6.3

v0.6.3.post1

v0.6.4

v0.6.4.post1

v0.6.5

v0.6.6

v0.6.6.post1

v0.7.0

v0.7.1

v0.7.2

v0.7.3

v0.8.0

v0.8.0rc1

v0.8.0rc2

v0.8.1

v0.8.2

v0.8.3

v0.8.3rc1

v0.8.4

v0.8.5

v0.8.5.post1

v0.9.0

v0.9.0.1

v0.9.1

v0.9.1rc1

v0.9.1rc2

v0.9.2

v0.9.2rc1

v0.9.2rc2

Mono Color

• 228d225c4e [DSV4] Guard megamoe flag with Pure TP (#41522) Yongye ZhuandRoger Wang 2026-05-02 19:41:40 -04:00
• 1c607d7b2c [DSV4] Guard megamoe flag with Pure TP (#41522) Yongye ZhuandGitHub 2026-05-02 19:41:40 -04:00
• 4f7309fcc0 [CI] Add ci-fetch-log.sh helper for Buildkite job logs (#41517) 3 people 2026-05-02 18:23:59 -04:00
• 0a9362d6ab Revert "[Build] Make bundled DeepGEMM wheel portable across Python versions" (#41512) Michael GoinandGitHub 2026-05-02 12:42:41 -04:00
• ae8c445789 Use -1 expert ID for padding and non-local tokens in DeepEP v2 deepep-v2-integration Tyler Michael SmithandClaude Opus 4.6 2026-05-02 11:51:27 -04:00
• cfd2573f23 [Build] Switch CUDA 13.0 wheel builds to PyTorch manylinux_2_28 base (#41416) 3 people 2026-05-02 08:51:28 -04:00
• a4debbd5bd Revert "Temporary disable persistent topk (#41442)" khluu 2026-05-02 00:52:06 -07:00
• 87494544b1 [Bugfix] Fix persistent_topk inter-CTA init race on RadixRowState (#41444) Yongye Zhuandkhluu 2026-05-01 17:42:52 -04:00
• c3ad791e1a [Bugfix][Gemma 4] Clamp soft-token estimate to max_soft_tokens (#40796) 5 people 2026-05-02 13:34:59 +07:00
• 8586369f61 Refactor Step3Text loading to use AutoWeightsLoader (#41492) Matthew SantiagoandGitHub 2026-05-02 01:22:14 -05:00
• 45a0b6d72d Fix local→global expert ID conversion in DeepEP v2 decode path Tyler Michael SmithandClaude Opus 4.6 2026-05-02 02:15:31 -04:00
• ae3b4deb8a [Doc] Add Codex usage example (#41358) ChaunceyandGitHub 2026-05-02 13:27:43 +08:00
• c293ccc58e [ROCm][Bugfix] Fix init-time bias dtype cast when gate.out_dtype is None (#41405) Rita BrugarolasandGitHub 2026-05-01 21:13:15 -07:00
• d58c42e19c [vLLM IR] 2/N fused_add_rms_norm and maybe_inplace overload (#36823) Luka GovedičandGitHub 2026-05-01 23:41:15 -04:00
• 3e49479c4b Limit concurrency on test_transcription_api_correctness.py (#41478) Ekagra RanjanandGitHub 2026-05-01 23:19:07 -04:00
• 964a4bc2a5 [MM][CG] Support ViT CG for Qwen2.5-VL (#40830) John CalderonandGitHub 2026-05-01 23:10:14 -04:00
• c408fdd663 [Fix] Sync gemma4 chat template from hf (#39570) FredericOdermattandGitHub 2026-05-02 05:06:54 +02:00
• 63e6293a38 Build Python from source instead of using deadsnakes PPA khluuandClaude Opus 4.6 2026-05-01 18:31:44 -07:00
• 7c36acf7fd Fix Python version regex quoting in Dockerfile releases/v0.20.1-python-from-source khluuandClaude Opus 4.6 2026-05-01 18:41:43 -07:00
• 9f876610de Build Python from source instead of using deadsnakes PPA khluuandClaude Opus 4.6 2026-05-01 18:31:44 -07:00
• e7c9359432 p khluu 2026-05-01 15:02:51 -07:00
• 5737770c6c Re-enable allreduce rms fusion for DP / PP (#41458) Andy LoandGitHub 2026-05-02 00:01:37 +01:00
• 747e379546 retry 10 times khluu 2026-05-01 15:54:13 -07:00
• 8ae4f9223e p khluu/release-registry-cache khluu 2026-05-01 15:02:51 -07:00
• 67c74e5980 p khluu/vllm-base-uv-python khluu 2026-05-01 14:47:49 -07:00
• 0c99629ede [Build] Make bundled DeepGEMM wheel portable across Python versions (#41476) Michael GoinandGitHub 2026-05-01 17:45:03 -04:00
• edd60ac93a [Bugfix] Fix persistent_topk inter-CTA init race on RadixRowState (#41444) Yongye ZhuandGitHub 2026-05-01 17:42:52 -04:00
• 455839ec66 p khluu 2026-05-01 14:28:31 -07:00
• 797cdc15f0 [CI] Use uv-installed standalone Python in vllm-base (drop deadsnakes) khluu/release-v0.20.1-uv-python khluuandClaude Opus 4.7 2026-05-01 13:42:32 -07:00
• 438f6548c1 [CI] Use deadsnakes mirror for non-Ubuntu24.04 release images khluuandClaude Opus 4.7 2026-05-01 13:25:05 -07:00
• 513b5f82dd [DSV4] Add knob to enable pre-attn gemm (#41443) Yongye Zhuandkhluu 2026-05-01 15:23:17 -04:00
• 135722d58d Revert "." khluu 2026-05-01 12:23:33 -07:00
• bcf5cac9fb [DSV4] Add knob to enable pre-attn gemm (#41443) Yongye ZhuandGitHub 2026-05-01 15:23:17 -04:00
• 7758daca7c [ROCm][Bugfix][GPTOSS]: fix input_ids and expert_map args for quark w4a8 gptoss (#41165) Rohan Potdarandkhluu 2026-04-29 18:39:01 -05:00
• 5f79634b3b [Perf] Intergrate Tile Kernels head_compute_mix_kernel for Deepseek-V4 (#41255) Isotr0pyandkhluu 2026-05-02 03:01:17 +08:00
• 05adde9955 [Bugfix] Pass reasoning parser kwargs to structured output (#41199) Bugen Zhaoandkhluu 2026-05-01 14:38:02 +08:00
• 9cdcda5fc2 [Bugfix] BailingMoeV2.5: rotate full qk_rope_head_dim in MLA RoPE (#41185) Jiangyun Zhuandkhluu 2026-04-29 18:08:55 +08:00
• a5d4a264dd [Bugfix ] fix bailing_moe_linear (#40859) JiangWeixiangandkhluu 2026-04-28 13:39:23 +08:00
• a9484dac7b [Perf] Intergrate Tile Kernels head_compute_mix_kernel for Deepseek-V4 (#41255) 3 people 2026-05-02 03:01:17 +08:00
• f3fef12350 [Attention] Abstract the MLA prefill backends and eliminate cuDNN (#32623) 5 people 2026-05-01 13:36:20 -04:00
• 51295793a2 [Model Runner V2] Add logprob_token_ids support (#40559) 3 people 2026-05-01 13:02:03 -04:00
• 3ccc1ff495 [Eval][CI] Add basic mrcr eval to tests/evals/ (#40164) Michael GoinandGitHub 2026-05-01 12:00:38 -04:00
• aeb92c37ab Merge branch 'main' into wentao-optimize-dcp-and-add-comm-func yewentao256 2026-05-01 15:54:28 +00:00
• 529c671e80 [ROCm][FEAT] AITER Fused Allreduce + RMSNorm (#37646) 6 people 2026-05-01 23:07:18 +08:00
• bc635fad23 [ROCm][Deepseek] dsv3.2 further optimization (#41217) 3 people 2026-05-01 22:06:00 +08:00
• c3e64696cd [Perf] Warmup forward_native sampler kernel (#41375) 3 people 2026-05-01 17:04:11 +03:00
• 4f7bde572a [Kernel] Pack output and LSE in DCP A2A (#41160) sungsoo haandGitHub 2026-05-01 06:01:17 -07:00
• 2fa1f8ec00 [kv_offload+HMA][13/N]: Enable HMA support (#41445) Or OzeriandGitHub 2026-05-01 14:30:03 +03:00
• 7075df79b3 [ROCm] Enable DBO (Dynamic Batch Optimization) on ROCm (#34726) raviguptaamdandGitHub 2026-05-01 02:18:30 -07:00
• 0dbaf9daad Refractor longcat loading to use AutoWeightsLoader (#41448) Yuyi AoandGitHub 2026-05-01 05:07:23 -04:00
• a3ec4a35f5 [Bugfix][Metrics] Fix RayPrometheusMetric.labels() returning shared labeled child (#40840) 4 people 2026-05-01 00:43:39 -07:00
• 32964e7700 [ROCm][CI] Upgraded UCX and RIXL (#41210) Andreas KaratzasandGitHub 2026-05-01 02:40:47 -05:00
• a07642667d [Bugfix] Pass reasoning parser kwargs to structured output (#41199) 3 people 2026-05-01 14:38:02 +08:00
• a5e0ea09ea . Yongye Zhuandkhluu 2026-05-01 04:17:52 +00:00
• c3868bbbe4 [compile] Add FlashInfer FP8 async TP fusion and preserve allreduce fusion ordering #27893 (#39505) baonudesifeizhaiandGitHub 2026-05-01 01:08:34 -04:00
• 947138b6c2 Add nvfp4 kv cache support (#40177) sychen52andGitHub 2026-04-30 21:55:16 -07:00
• 28a1a85d89 [DSV4] Add BF16 and MXFP8 A2A support for flashinfer a2a one sided (#40960) Yongye Zhuandkhluu 2026-04-30 18:33:12 -04:00
• 82ccd9f3b2 [Bugfix] Fix persistent_topk cooperative deadlock at TopK=1024 (#41189) Yongye Zhuandkhluu 2026-04-30 00:03:45 -04:00
• d9f3961abb [DSv4] Use cvt PTX for FP32->FP4 conversion (#41015) Thien Tranandkhluu 2026-04-30 06:16:40 +07:00
• 4fee1390f4 [Bugfix] Fix repeated DSv4 RoPE cache initialization (#41148) Jee Jee Liandkhluu 2026-04-29 20:29:55 +08:00
• bf590b6a24 [Bugfix] DSV32/V4 add missing type conversion for non-streaming tool calls (#41198) Chaunceyandkhluu 2026-04-29 17:55:07 +08:00
• 66df2c03f5 [Bugfix] Fix max_num_batched_token not captured in cuda graph (#40734) Wei Zhaoandkhluu 2026-04-29 00:33:06 -04:00
• 17896d817b [Bugfix] fix inductor error for dpsk v4 (#41135) Jiangyun Zhuandkhluu 2026-04-29 12:18:46 +08:00
• 1a131a066c [Bugfix] Fix Deepseek V4 import error due to AOT compile cache loading (#41090) Wei Zhaoandkhluu 2026-04-29 00:08:16 -04:00
• 6fffb0b8c8 [DSV4] Align aux stream API with DeepseekV4DecoderLayer (#41171) qizixiandkhluu 2026-04-28 17:22:03 -07:00
• a076bd1e67 [Core] Account for num_gpu_blocks_override in max_model_len checks (#41069) Nick Hillandkhluu 2026-04-28 15:44:09 -07:00
• c244b6fc83 [DSV4] Enable Multi-stream for Pre-Attn GEMM (#41061) Yongye Zhuandkhluu 2026-04-28 12:08:55 -04:00
• 15cc87d2fe [Model][DSV4] Support base model (#41006) Jee Jee Liandkhluu 2026-04-28 08:16:55 +08:00
• c68fa39ced Auto-disable expandable_segments around cumem memory pool (#40812) youkaichaoandkhluu 2026-04-27 09:37:22 +08:00
• 941fb50835 [kv_offload+HMA][12/N]: Scheduler-side support for sliding window groups (#41228) Or OzeriandGitHub 2026-05-01 06:59:17 +03:00
• 6b6ac6c3c7 [Kernel][MoE] Support GELU on TRT-LLM NvFP4 fused MoE for Gemma4 (#41050) 3 people 2026-04-30 20:37:43 -07:00
• 05ebca5250 Temporary disable persistent topk (#41442) Yongye ZhuandGitHub 2026-04-30 23:37:19 -04:00
• b542bdf7fb [Bugfix] Disable FlashInfer CUTLASS MoE on SM110 (Jetson Thor AGX) (#40808) Stefano CastagnettaandGitHub 2026-05-01 05:08:49 +02:00
• 0f94cbef14 Fix FP8 padding: use torch.where instead of masked_fill_ Tyler Michael SmithandClaude Opus 4.6 2026-04-30 22:54:59 -04:00
• ed29278a33 Split DeepEP v2 into prefill/decode modes for memory vs cudagraph Tyler Michael SmithandClaude Opus 4.6 2026-04-30 22:19:23 -04:00
• 415a879899 [KV Offload] Use Collection instead of Sequence/Iterable for OffloadingManager key parameters (#41361) Ronen SchafferandGitHub 2026-05-01 05:18:38 +03:00
• 7198940b39 [Model] Add Moondream3 model support(only query and caption skills) (#32325) Dong WandGitHub 2026-04-30 19:06:48 -07:00
• 14043dfecd feat: Enable prompt_embeds Content Part Support in vLLM Chat Completions API (#40720) 3 people 2026-04-30 22:05:55 -04:00
• 1adaa5056b [ROCm][CI] Add ROCm score absolute tolerance floor (#41341) Andreas KaratzasandGitHub 2026-04-30 20:59:35 -05:00
• 4d5c89295b (bugfix): block_size check for flex attn (#41363) 3 people 2026-05-01 09:59:26 +08:00
• 56a8767d94 Update DeepEP v2 docstring with cudagraph design rationale Tyler Michael SmithandClaude Opus 4.6 2026-04-30 21:55:14 -04:00
• 5a766bf3de Refactor DeepEP v2 for cudagraph compatibility and simplified DBO Tyler Michael SmithandClaude Opus 4.6 2026-04-30 21:51:16 -04:00
• dd5506a157 [Core] Simplify handling of scheduler_reserve_full_isl option (#41064) Nick HillandGitHub 2026-04-30 18:10:00 -07:00
• a3c83ff2fd Faster per-token fp8 group quant packed kernel for blackwell (#41326) 3 people 2026-04-30 21:09:55 -04:00
• 9c61864bf8 [DeepSeek] Use torch.mm for bf16xbf16->fp32 gemm (#41300) Woosuk KwonandGitHub 2026-04-30 16:28:57 -07:00
• 71725f6730 [Bugfix] Fix RoutedExpertsCapturer for Gemma 4 MoE (top_k_experts) (#41401) Tran LeandGitHub 2026-04-30 16:19:59 -07:00
• b4806c8ee1 [DSV4] Add BF16 and MXFP8 A2A support for flashinfer a2a one sided (#40960) 4 people 2026-04-30 18:33:12 -04:00
• 526927be94 [Model Runner v2] Fix v2 compile counter num_gpu_runner_capture_triggers and num_cudagraph_captured (#41285) Wentao YeandGitHub 2026-04-30 18:20:11 -04:00
• 75a4c166f2 Fix typo in log message for indexer cache (#41419) Michael GoinandGitHub 2026-04-30 18:02:14 -04:00
• 2917d6363a [NVFP4][Hopper/AMD Instinct] Add Triton kernels for NVFP4 dequantization and QDQ emulation (#40033) 3 people 2026-04-30 23:35:48 +02:00
• 77230471c0 remove torch 2.9, 2.10 workround wentao-remove-dead-code yewentao256 2026-04-30 21:10:50 +00:00
• 2c1cf4e831 [CI] Trigger perf-eval after nightly Docker publish khluuandCodex 2026-04-30 13:29:25 -07:00
• efb4cdf2b8 [CI/Build] Skip Prithvi/Terratorch model-registry tests when terratorch is missing (#41389) Stefano CastagnettaandGitHub 2026-04-30 21:47:55 +02:00
• 92a7c121b6 [CI] Add MTP coverage: Qwen3.5 correctness + no-sync spec decode (#40472) 3 people 2026-04-30 21:24:09 +02:00
• 307b17ce33 [DSV4] Avoid redundant dtype conversion. (#41374) Jee Jee LiandGitHub 2026-05-01 00:57:27 +08:00
• 3ca6ca210f xpu docker: pin oneAPI to 2025.3 and avoid unintended 2026 upgrade (#41380) 3 people 2026-05-01 00:02:23 +08:00
• 10558f5f46 [CI/Build] Skip terratorch + torchgeo while PyPI has lightning quarantined (#41377) Stefano CastagnettaandGitHub 2026-04-30 16:59:07 +02:00
• 121dbe7a22 [ROCm] ROCm DeepEP API updated to latest (#39721) 4 people 2026-04-30 09:46:59 -05:00
• f03d82efdd [UX][Bugfix] Fix OOM by setting PyTorch max_split_size_mb during model loading (#41268) Matthew BonanniandGitHub 2026-04-30 10:46:54 -04:00
• a7fb008510 [EPLB] Optimize memory overhead in Nixl communicator (#40013) 4 people 2026-04-30 16:46:49 +02:00