Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• 916e56c05c [QeRL] Add warnings for extra memory buffering (#40309) Kyle SayersandFlora Feng 2026-04-29 00:06:54 -04:00
• a085b5257d [Docs] [QeRL] Layerwise Reloading Documentation (#40317) Kyle SayersandFlora Feng 2026-04-29 00:06:38 -04:00
• 7fd05e05ae uncomment flex backend for batch invariant mode (#40842) liangel-02 2026-04-29 00:05:14 -04:00
• 99255f3cb5 [UX] Allow enable/disable model weights loading tracking by config (#41086) Isotr0pyandCopilot 2026-04-29 12:04:49 +08:00
• 75a7cf2c10 [CI] De-flake test_chat_completion_n_parameter_non_streaming (#41147) haosdent 2026-04-29 11:23:59 +08:00
• 4b95e9cec4 [CI] Return HTTP 400 for unsupported chat content part type (#41121) haosdent 2026-04-29 10:23:26 +08:00
• 856b15c62c [CI][AMD][BugFix] Patch has_flashinfer decorator for test_select_rocm_aiter_backend (#41072) rasmith 2026-04-28 21:12:17 -05:00
• 6fb3f7b46b [DSV4] Align aux stream API with DeepseekV4DecoderLayer (#41171) qizixi 2026-04-28 17:22:03 -07:00
• d109eacd05 [Bugfix][ROCm] Fix gemm_a4w4 call to use updated AITER API signature (#40754) chelnnexy 2026-04-28 19:04:53 -05:00
• e68fa1b90a [Core] Account for num_gpu_blocks_override in max_model_len checks (#41069) Nick Hill 2026-04-28 15:44:09 -07:00
• f05f3664c3 [Doc] Add missing API endpoints to security documentation (#40532) Russell Bryant 2026-04-28 17:53:19 -04:00
• e9f8f31e9a [FEATURE] Add EagleMistralForCausalLM (#41024) Julien Denize 2026-04-28 21:22:20 +02:00
• de3fe8dc62 [Bugfix] release KV blocks for skipped P-ranks to prevent invalid KV errors and timeouts when P_tp > D_tp and MLA (#40449) yangrzandRoger Wang 2026-04-29 02:38:43 +08:00
• 0899f436aa [New Model] Laguna XS.2 implementation (#41129) Joe RowellandRobert Shaw 2026-04-28 20:23:00 +02:00
• 358a755e43 [CI][AMD][BugFix] Update request URL in test_moriio_connector to match vllm-router compatibility changes (#41076) rasmith 2026-04-28 13:14:59 -05:00
• a60883644b [Build] Defer flashinfer cubin download to avoid ~2.5 GB (decompressed) layer duplication (#41134) Benoit Tigeot 2026-04-28 19:27:18 +02:00
• 5aa371dc8e [DSV4] Enable Multi-stream for Pre-Attn GEMM (#41061) Yongye Zhu 2026-04-28 12:08:55 -04:00
• de3da0b97c Add tuned triton fused_moe configs on H100 for gpt-oss (#39904) zhangxin81 2026-04-28 18:38:48 +08:00
• 9e92de51c6 [Bugfix] Exclude numa_bind fields from ParallelConfig DP hash (#41098) Roy Wang 2026-04-28 15:52:54 +08:00
• bde0efdbb7 [Bugfix][Granite4Vision] Fix deepstack buffer causing decode slowdown in compiled mode (#40917) artem-spectorandartemspector 2026-04-28 10:43:30 +03:00
• ea74f701db Bugfix: fix SpecBench sample argument error (#40927) zhrrr 2026-04-28 15:33:49 +08:00
• a8208e6a81 [Examples] Resettle features examples. (#40995) wang.yuqi 2026-04-28 15:33:41 +08:00
• 76c9cccc36 [Core] Fix redundant None append in StepPool.forward for chunked prefill (#41049) anthonsu 2026-04-27 23:42:47 -07:00
• ed57f77192 [Bugfix ] fix bailing_moe_linear (#40859) JiangWeixiang 2026-04-28 13:39:23 +08:00
• 7a1eb8ac2e [Model] update for mimo v25 (#41029) 3 people 2026-04-28 12:52:54 +08:00
• c2e88a281c [Bugfix] Fix broken example opeanai client (#41088) Isotr0py 2026-04-28 12:43:04 +08:00
• fd74c90d9c [Attention][Spec Decode] Allow independent drafter attention backend selection (#39930) Matthew Bonanni 2026-04-27 22:38:09 -04:00
• 146f44b77d [Frontend]Responses API supports Tool/Function calling with streaming with required (#40700) Chauncey 2026-04-28 10:36:58 +08:00
• 0d4f714208 [Bugfix] Remove tokenizer encode/decode calls from Olmo3 reasoning parser (#40855) yzong-rhandFlora Feng 2026-04-27 22:36:54 -04:00
• 03aeed802f [Test] Fix test_dynamic_shapes_compilation for torch 2.12 (#40743) Angela Yi 2026-04-27 17:51:15 -07:00
• 2c8b76c5cb [Model][DSV4] Support base model (#41006) Jee Jee Li 2026-04-28 08:16:55 +08:00
• 407b34be26 [xpu] bump up vllm-xpu-kernel v0.1.7 (#41019) Kunshang Ji 2026-04-28 08:04:54 +08:00
• 4c7c69b4e0 [Model Runner V2] Skip attention metadata rebuild before draft prefill (#40410) Giancarlo Delfin 2026-04-27 15:38:05 -07:00
• 5e2c37facd [ROCm][CI] Add missing quantization methods and fix online quant test failures (#39801) Andreas Karatzas 2026-04-27 15:08:57 -05:00
• c8bbe05189 [Perf] Update TRTLLM supported MoE routing methods (#39141) 4 people 2026-04-27 14:16:22 -04:00
• 6232fb4b66 [Docker] Install numactl CLI in CUDA runtime image (#41032) 3 people 2026-04-27 10:58:06 -07:00
• 2c06cf3486 [Bugfix] use served_model_name for multimodal error message (#41003) Moritz Sanft 2026-04-27 17:22:35 +02:00
• e6f710a87f Deprecate support for Transformers v4 (#40389) Harry Mellorandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-04-27 16:19:57 +01:00
• c245d35ff4 [Model] Add MiMo-V2.5 support (#40967) 7 people 2026-04-27 21:26:51 +08:00
• f8ac0c7cf0 [Bugfix] Fix k_norm weight sharding in MiniMaxM2Attention when total_num_kv_heads < tp_size (#38191) Xiaoshuang Wangandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-04-27 20:57:13 +08:00
• ebf862c351 Add system_fingerprint field to OpenAI-compatible API responses (#40537) Simon MoandClaude 2026-04-27 01:17:52 -07:00
• 8d8062d0a7 [Examples] Resettle generate examples. (#36464) wang.yuqi 2026-04-27 15:48:37 +08:00
• 985961345a [Bugfix] Install libcublas-dev in Dockerfile for FlashInfer CuTe DSL JIT (#39855) Roy WangandRoger Wang 2026-04-27 15:47:39 +08:00
• 706a04d34b [DSV4] Add silu clamp limit to shared expert (#40950) Yongye Zhu 2026-04-27 03:37:43 -04:00
• 22631f80a0 [Bugfix] Remove invalid deepstack boundary check for Qwen3-VL (#40932) Isotr0py 2026-04-27 15:27:06 +08:00
• 2cc008e7b4 [Attention][TurboQuant] Share dequant buffers, eliminate float16_copy (#40941) Bhoomit 2026-04-26 22:48:36 -07:00
• 5d5c776444 [Perf] FP8 FlashInfer Attn for ViT (#38065) Zhanda ZhuandYubo Gao 2026-04-27 06:44:15 +01:00
• 592ae6805c Cutlass W4A16 (Machete) Tests (#35450) ojhaanshika 2026-04-26 22:15:29 -07:00
• 7b1bc0a3eb [Bugfix] Cap SWA/chunked-local runtime admission to startup pool-sizing bound (#40946) 3 people 2026-04-26 21:33:13 -07:00
• c0879d9483 [Tests] Gate Isaac under Transformers v5 (#40907) Silu Panda 2026-04-26 19:26:51 -07:00
• f5f9878514 [Model Runner V2] Fix rejection sampling acceptance rate gap vs MRV1 (#40651) Giancarlo Delfin 2026-04-26 19:12:08 -07:00
• 2ce95a761b Auto-disable expandable_segments around cumem memory pool (#40812) 3 people 2026-04-27 09:37:22 +08:00
• 4d51588e23 [Feat] DeepSeek V4 Rebased (#40860) +717 people 2026-04-26 18:31:08 -07:00
• 32e45636e3 [torch.compile]: Disable Sequence Parallelism (SP) for piecewise compilation (#38373) 6 people 2026-04-27 01:44:42 +08:00
• b39c266dae [KV Offload] Offload all KV blocks when doing prefill in P/D (#40346) omerpaz95andOr Ozeri 2026-04-26 15:06:01 +03:00
• 9558f43903 [Bugfix] Size FlashInfer NVLink MNNVL workspace to EP group (#40893) Dao007forever 2026-04-26 01:26:34 -07:00
• 8cd174fa35 [LoRA] MoE LoRA Refactor (#40338) Jee Jee Li 2026-04-26 09:55:19 +08:00
• c798593f0d [Bugfix] Fix the DSML token leakage in DSV4/3.2 (#40806) 3 people 2026-04-26 08:58:50 +08:00
• 12a3f6454b [Bugfix][MoE] Only unpad routed output before shared expert add or routed output transform (#40865) Netanel HaberandJee Jee Li 2026-04-25 23:50:12 +03:00
• 60cd878a3b [kv_offload+HMA][11/N]: Support store with multiple KV groups (#39403) Or Ozeri 2026-04-25 20:00:46 +03:00
• 1e9f19ca3f [CI][AMD]BugFix] Fix deadlock occuring in test_moe_layer (#40767) rasmith 2026-04-25 08:34:14 -05:00
• 6646c0c7e0 [Opt] Optimize deepstack buffer handling for multimodal Qwen3 models (#40145) labAxiaoming 2026-04-25 21:04:26 +08:00
• 95995bbef8 [ROCm][Engine] Fix GPU memory leaks in engine shutdown and test workaround for async KV prefix cache reset (#38503) Andreas Karatzas 2026-04-25 00:25:20 -05:00
• 07351e0883 [Feature] Warm up readonly multimodal processor during renderer startup (#40797) Chenguang ZhengandOpenAI Codex 2026-04-25 11:57:41 +08:00
• 428b988c98 [ROCm][CI] Fix trust_remote_code AttributeError in EAGLE3 acceptance length test (#40306) Andreas Karatzas 2026-04-24 21:59:31 -05:00
• e54894fc85 [ROCm][CI] Fix TestSiluMulGroupFp8QuantModel after W8A8 block linear refactor (#39799) Andreas Karatzas 2026-04-24 21:20:59 -05:00
• bc2ae5a3d6 [Test] Increase qwen2_vl num_logprobs to fix torch 2.12 update (#40818) Angela Yi 2026-04-24 17:59:20 -07:00
• a474da2813 [Refactor] Remove unused dead code (#40640) Wentao Ye 2026-04-24 19:28:18 -04:00
• ce6a199ecc [BE][Bugfix] Respect TORCH_COMPILE_DISABLE env var at the vLLM config level for torch 2.12 (#40715) Lucas Kabela 2026-04-24 16:25:03 -07:00
• f88763efc3 [Bugfix] add seq_lens_cpu_upper_bound to CommonAttentionMetadata in mla_runner.py (#40844) Ignacio Sica 2026-04-24 20:13:52 -03:00
• 333529deae [EPLB] Fix replica selection bias in fused_moe router (#40810) Artem Perevedentsev 2026-04-25 01:06:41 +03:00
• 8825608205 [Bugfix][CI] Fix wrong residual shape in TestFusedAddRMSNorm.example_inputs that causes flaky test (#40629) Zhang Jian 2026-04-25 04:40:07 +08:00
• 095d2f87e8 [Bug] Fix GLM-5.1 running error on ROCm platform (#40763) qli88 2026-04-24 14:54:40 -05:00
• 21792520e7 [Build] Add Python 3.14 to supported version list. (#34770) Neil SchemenauerandSimon Mo 2026-04-24 10:24:05 -07:00
• 5e11b40365 [Frontend] Delegate to vLLM Omni When --omni Passed (#40744) Alex Brooks 2026-04-24 10:30:00 -06:00
• f768b4473e [Docs] Add docs for context extension using the yarn method (#37430) labAxiaomingandCyrus Leung 2026-04-24 23:26:09 +08:00
• 914d0464c1 [Refactor] Unify 2D/3D kernels in triton_unified_attention (#40631) JartX 2026-04-24 17:18:06 +02:00
• 9f771b3ab9 [Quantization] add humming quantization kernel (#34556) Jinzhen Lin 2026-04-24 21:29:44 +08:00
• c9d3c6e6af fused_moe: treat NIXL EP as batched experts (#40412) Itay Alroy 2026-04-24 16:05:31 +03:00
• 51adca74e6 [kv_offload+HMA][9/N]: Support lookup with multiple KV groups (#39401) Or Ozeri 2026-04-24 15:32:29 +03:00
• e8eb0490ce [Bugfix][MoE] Unpad routed output before shared expert add [Fixes #35949] (#40794) Netanel Haber 2026-04-24 14:53:23 +03:00
• e8ee2a78db [Attention] use diff kv backend for mimo v2 flash (#40045) Jiangyun Zhu 2026-04-24 19:25:55 +08:00
• 2ec18f5df4 [Bugfix][Parser] Fix Mistral tool parser for HF tokenizers (#39294) 4 people 2026-04-24 13:01:56 +02:00
• 6dec49f27e [Build] Bump CUDA to 13.0.2 to match PyTorch 2.11.0 (#40669) Dmitry Tokarev 2026-04-24 06:27:11 -04:00
• b5587e1013 [CI/Build] Add e2e test for ViT CUDA graph (#40780) Shanshan Shen 2026-04-24 18:12:14 +08:00
• 9ad5abe772 Fix Nano Nemotron VL static image inputs (#40724) milesial 2026-04-24 02:18:55 -07:00
• 7d3195ea9f [Bugfix] Fix IMA in DSA + MTP (#40772) Woosuk Kwon 2026-04-24 01:40:20 -07:00
• 512f522192 [Model] Gemma4: add bidirectional vision attention for sliding layers with window guard (#40534) 4 people 2026-04-24 01:27:46 -07:00
• 4c34b2f6fc [XPU] Enable torch.compile for XPU GDN attention (#39466) Yuwen ZhouandKunshang Ji 2026-04-24 16:26:16 +08:00
• cf8a613a87 Support only half types for concat_mla_q kernel (#37892) Xin Yang 2026-04-23 23:51:05 -07:00
• 01acf96c6f [XPU][CI] Fix Docker cleanup races on Intel CI runners (#40761) xiangdong 2026-04-24 14:08:45 +08:00
• 079a4cf399 [MoE] Move cutlass moe to fused_moe/experts/ (#40574) Jackmin801andClaude 2026-04-23 23:05:49 -07:00
• 9744b699ba [Deprecate] Deprecate LLM.reward offline api, use LLM.encode instead. (#40688) 3 people 2026-04-24 13:37:50 +08:00
• c662b4359e [Bugfix] Avoid mutating chat_template_kwargs in HYV3ReasoningParser initialization (#40713) Bugen Zhaoandgemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> 2026-04-24 13:08:58 +08:00
• 100c7b65e7 [Platform] Fix RISC-V platform detection (lscpu parsing + non-NUMA meminfo) (#40427) lyd1992 2026-04-24 12:33:05 +08:00
• 56bdf85e10 [Feature] Avoid eager import of the "mistral_common" package. (#40043) Neil Schemenauer 2026-04-23 19:49:16 -07:00
• eba73068ea [Doc] fix capitalization consistency in README (vLLM, Hugging Face) (#40729) Vinayak Kumar 2026-04-24 07:53:54 +05:30
• e9f331d72e [MRV2] Ensure warmup covers prefill path (#40746) Nick Hill 2026-04-23 18:33:26 -07:00
• c9bf77df92 [BUG]: fix HF tokenizer concurrent borrow in tool parsers (#40059) 3 people 2026-04-23 21:20:30 -04:00
• 3041344287 [Misc] Added curl retries in install_python_libraries.sh (#36700) Dmitry Tokarevandgemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> 2026-04-23 21:19:30 -04:00