Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86

Mono Color

• 61fbe0b865 [Mini-VLLM] Remove unreferenced definitions found by static scan sm86 MOVZXandHinata 2026-10-19 21:39:56 +07:00
• 5aa2ac64f8 [Mini-VLLM] Remove LoRA, plugins, speech-to-text and scale-out surfaces MOVZXandHinata 2026-10-19 21:06:19 +07:00
• 5f3c7bd8ab [Mini-VLLM] Remove the ROCm device path, debug routes and arch-gated tests MOVZXandHinata 2026-10-19 19:31:07 +07:00
• 2066d3143f [Mini-VLLM] Cut non-OpenAI surfaces, dead kernels, Mistral stack MOVZXandHinata 2026-10-19 17:10:56 +07:00
• bd70a61c20 [Mini-VLLM] Remove orphan subsystems: snapshot, CPU workers, dead decoders MOVZXandHinata 2026-10-19 09:27:06 +07:00
• 2b544b7ba0 [Mini-VLLM] Cut dead subsystems across serving, connectors, parsers MOVZXandHinata 2026-10-19 02:19:03 +07:00
• 49d708e158 Merge branch 'main' into patches patches MOVZX 2026-10-18 18:09:11 +00:00
• 54a396586e [Mini-VLLM] Repair test suite after SM86 cut MOVZXandHinata 2026-10-19 01:06:30 +07:00
• 0152af1336 [Mini-VLLM] Remove upstream CI, docs, and lint surfaces MOVZXandHinata 2026-10-19 00:41:57 +07:00
• e3af324d8a [Mini-VLLM] Drop upstream model recipes directory MOVZXandHinata 2026-10-19 00:38:54 +07:00
• 8dfaffb700 [Bugfix] Restore xpu_communicator for CUDA graph capture MOVZXandHinata 2026-10-19 00:33:19 +07:00
• 81074e94a3 [Mini-vLLM] Cut dead subsystems: models, entrypoints, backends, utils, tests MOVZXandHinata 2026-10-19 00:20:25 +07:00
• 6a95f62e2f [Bugfix] Load checkpoint KV-cache scales in the Qwen3.5 family MOVZXandHinata 2026-10-18 17:18:47 +00:00
• 6cf86cd95f [Bugfix] Load checkpoint KV-cache scales in the Qwen3.5 family MOVZXandHinata 2026-10-18 17:18:47 +00:00
• ff01a6b14f [Bugfix][Frontend] Stop beam search on generation config EOS tokens (#62072) main alpandmottopanikeiku 2026-10-18 09:49:42 -07:00
• f36850e389 [Mini-VLLM] Restore FlashInfer backend for FP8 KV & only probe quantization overrides that exist MOVZXandHinata 2026-10-18 23:17:51 +07:00
• fe8f2140ee [Doc] Document /inference/v1/generate request/response fields and DP rank header (#62305) aoshen02andClaude Opus 5.5 2026-10-18 23:26:36 +08:00
• e80b6c642f [ROCm][DSv4] Overlap the unfused shared expert with AITER MegaMoE (#61819) 3 people 2026-10-18 22:39:38 +08:00
• 371745ebd6 [Doc][RL] Document routed experts return for MoE routing replay (R3) (#62302) aoshen02andClaude Opus 5.5 2026-10-18 21:11:45 +08:00
• e61dcb9143 [Mini-vLLM] SM86-only cut: models, backends, Rust, tests MOVZXandHinata 2026-10-18 19:19:56 +07:00
• 072c0886d5 [Build] Keep plain CUDA arch when its suffixed variant is also requested (#56549) 3 people 2026-10-18 08:13:17 -04:00
• 6b24947628 [CI/Build] Build DeepEP for SM103 (B300/GB300) in the Docker image (#61829) 3 people 2026-10-18 14:12:34 +02:00
• 494663940b [Qwen4Exp] Clean up HC and PLE implementations for Qwen3.8-Flash-Next (#61454) Harry HuangandThien Tran 2026-10-18 19:38:21 +08:00
• 3c13653738 [Frontend] Score with the chat template declared by Sentence Transformers CrossEncoders (#60877) Rémi Delhaye 2026-10-18 13:30:12 +02:00
• 953fa442fa [CI] Per-test timeout for Language Models (Extended Pooling) (#62279) Kevin H. LuuandClaude Opus 5.5 2026-10-18 04:00:16 -07:00
• 3f1b04e045 [Bugfix] Fix Opencv video backend frames count missing for WebM video inputs (#61759) Isotr0py 2026-10-18 18:37:39 +08:00
• 2365177ddd [Doc][RL] Document the weight-transfer HTTP protocol for trainers without vLLM (#62267) aoshen02andClaude Opus 5.5 2026-10-18 17:26:56 +08:00
• c5ab42bc0f Merge branch 'main' into patches MOVZX 2026-10-18 16:04:36 +07:00
• 1b2a666391 [Doc][Recipe] Fix for redhat fp8 xeon support issue (#62245) Louie Tsai 2026-10-18 01:46:56 -07:00
• 83c9a547bf [CI][Release] Fix release wheel builds: stop excluding tracked recipes/ from Docker context (#62221) 3 people 2026-10-18 00:48:00 -07:00
• b510857d37 [Bugfix][Build] Build fused GDN MTP decode for SM80-SM90 with CUDA < 13 (#60896) 4 people 2026-10-18 15:14:06 +08:00
• 7c9042c05a [Bugfix][MRV2] Fix speculative warmup deadlock with pipeline parallelism (#62129) 3 people 2026-10-18 14:51:32 +08:00
• da9134bc89 [ROCm][CI] Give the MoE parallel-launch watchdog room for AITER JIT builds (#62161) 3 people 2026-10-17 21:49:27 -05:00
• b2456c622b [ROCm][CI] Make ROCm 10.0 the default image; keep ROCm 7.2 as -rocm72 (#58761) 5 people 2026-10-17 21:43:59 -05:00
• 03ee2e134d [CI][ROCm] Fix order-dependent test_rocm_aiter_sampler_defers_import (CPU tensors hit Triton) (#62213) 3 people 2026-10-17 18:10:08 -07:00
• 9c778e9ff8 [Bugfix][EPLB] Wait for pending expert buffer copies before NIXL READs (#62081) djramicandAndreas Karatzas 2026-10-18 01:18:02 +02:00
• 7879f3687b [NIXL] Drop unreachable producer-side notif counting in push mode (#62210) 3 people 2026-10-17 14:24:35 -07:00
• 255f14599e [Bugfix][Rust Frontend] Fix macOS build: read inherited listener family from its address (#62177) Kevin H. LuuandClaude Opus 5.5 2026-10-17 11:56:28 -07:00
• 03e8300d2c [Bugfix] Enable inductor batch-invariant mode under VLLM_BATCH_INVARIANT (#61118) 4 people 2026-10-17 11:53:06 -07:00
• 5c1c903501 [Perf][Parser] Stream Qwen3 tool argument values incrementally (12 KB call: 2.1 s to 11 ms) (#62027) Afonso RibeiroandFlora Feng 2026-10-17 19:49:06 +01:00
• bafb263664 [Bugfix][Benchmarks] Name the real cause when the /tokenize alignment probe fails (#59888) Tyrone 2026-10-17 20:47:52 +02:00
• 0c5dd18e7a [MRV2] Collect KV transfer results before the drafter runs (#62041) Nick HillandClaude Opus 5.5 2026-10-17 10:27:57 -07:00
• 763e3f7c91 [Bugfix][Spec Decode] Disable dynamic speculative decoding for fixed-K proposers instead of crashing (#46399) 3 people 2026-10-18 00:41:04 +08:00
• e744eacf68 [Rust Frontend] Validate bad_words limits during request lowering (#61933) AlecandClaude Opus 5.5 2026-10-17 09:10:40 -07:00
• c0b6f74bee [Benchmark] Add per-case kernel profiles to vllm bench forward (#61692) 3 people 2026-10-17 08:27:11 -07:00
• c61a1e73c6 [Feature] Skip MLA weight absorption when weights are pre-processed (#57320) siyuandXun Sun 2026-10-17 23:10:04 +08:00
• 237d99b4af Revert "[Docs] Add train-inference consistency agent skill" (#61551) (#62186) aoshen02andClaude Opus 5.5 2026-10-17 21:54:30 +08:00
• 2d8efee9a7 [Docs] Add train-inference consistency agent skill (#61551) aoshen02andClaude Opus 5.5 2026-10-17 20:49:50 +08:00
• 21e09ea959 [Doc] Sync DeepSeek-V4.1-Flash recipe with vllm-project/recipes#1136, #1137 (#62179) Roy WangandClaude Opus 5.5 2026-10-17 20:27:52 +08:00
• 837e87e1c5 [Feature][LoRA][MoE] Expert LoRA for FlashInfer TRTLLM MXFP4 experts (#61979) 3 people 2026-10-17 20:17:56 +08:00
• 4e964b2623 [XPU] Fix Event init failure w/ external (#61743) liuzhenweiandKunshang Ji 2026-10-17 19:10:39 +08:00
• ec8b20fbf0 [Bugfix] Fix Gemma 4 encoder CUDA graph capture OOM at startup (#61365) Vadim GimpelsonandClaude 2026-10-17 15:04:49 +04:00
• 2b85d9cb65 [Perf][Bugfix] Prefer MNNVL for single-node FlashInfer allreduce again (#61989) 3 people 2026-10-17 02:55:28 -07:00
• 026f6ab268 [Model] DeepSeek-V4/V4.1 VL: stream language-model weights instead of sorting the checkpoint (#58857) 3 people 2026-10-17 02:37:56 -07:00
• b515648641 [Doc] Host model recipes in recipes/ (#61302) Roy WangandClaude Opus 5.5 2026-10-17 17:14:36 +08:00
• f98ae6a9fd [Bugfix][Core] Keep resumable sessions that stop while preempted (#58778) 3 people 2026-10-17 02:08:14 -07:00
• 2067297b2f [ROCm][Perf] Use AITER tuned configs to select a4w4 gemm kernel instead of env var (#61278) 4 people 2026-10-17 04:00:28 -05:00
• 3bf25c5097 [CI][Bugfix] Ignore AMD CI webhook no-op builds in /amd-ci run and retry (#62087) Aarushi JainandAndreas Karatzas 2026-10-17 03:54:31 -05:00
• e56b1603af [Kernel]Add norm fused kernel for Step-5 Preview and Muse Spark Open (#62132) Jee Jee Li 2026-10-17 16:53:00 +08:00
• eb8de8d1a1 [Bugfix][OpenAI] Respect allowed_token_ids in beam search (#47630) Ting SUN 2026-10-17 16:50:40 +08:00
• b05ede5616 [Bugfix][Frontend] Preserve token IDs in completion derender responses (#62114) dskwe 2026-10-17 16:27:48 +08:00
• 61bb590877 [Bugfix][Spec Decode] Clear stale shared block-table rows (#61174) 3 people 2026-10-17 16:25:02 +08:00
• c3de0b37e2 [Core][Spec Decode] Replace EAGLE group heuristics with explicit draft-layer tags (#61455) 3 people 2026-10-17 16:22:33 +08:00
• e0fe5df731 Merge branch 'main' into patches MOVZX 2026-10-17 08:06:56 +00:00
• 08a65e287e [Bugfix][Kernel] hadacore: don't load past the input on a partial last warp (#61944) 4 people 2026-10-17 00:52:17 -07:00
• ad2688ab30 [Bugfix] Validate draft-token capacity in forward benchmark (#61809) znzrybandWoosuk Kwon 2026-10-17 15:51:36 +08:00
• 5be4266f5c [Bugfix][Frontend] Preserve Cohere and Anthropic streaming errors (#61947) JovanandOpenAI Codex 2026-10-17 14:10:16 +08:00
• 90d6d871d6 [Attention] Add CuTe DSL backend selection for MLA context prefill (#62069) Wei Zhao 2026-10-17 01:21:55 -04:00
• d74f4180cd [Bugfix][Frontend] Detokenize echoed prompts after truncation (#61876) Cy ChengandCy 2026-10-16 22:20:38 -07:00
• 39983a5e7a [Kernel] Add FP32 router GEMM support for Step-5 Preview and Muse Spark Open (#62126) Jee Jee Li 2026-10-17 13:17:55 +08:00
• 41b8c15809 [Feature][MoRIIO] Support hybrid WRITE transfers with DSpark (#60544) 4 people 2026-10-17 12:57:58 +08:00
• a0e7137c4b [CI][ROCm][Bugfix] DeepSeek V4.1: don't pass expert intermediate padding to AITER (#61960) Micah WilliamsonandCursor 2026-10-16 23:25:39 -05:00
• 7b6310c109 [Perf][Sampler] Speed up the small-batch split top-p pipeline (#61996) 4 people 2026-10-17 12:22:11 +08:00
• 556ac15eb9 [Bugfix] Remove extra args from CPU's compute slot mapping kernel (#61992) Fadi Arafehandvllm-agent 2026-10-17 05:11:33 +01:00
• dcd07967ca [CPU] Keep the hybrid block size across the engine core's config rerun (#61805) Gabriel WuandClaude Opus 5.5 2026-10-17 12:06:03 +08:00
• ce49c47cbb [Bugfix][Sampler] Carry the boundary tie count across blocks in the split top-p mask kernel (#62046) 3 people 2026-10-17 07:56:57 +05:30
• ce8853d27a [Bugfix][KV Offload] Allow explicit SimpleCPUOffload block counts (#61674) Sherif Walyandvllm-agent 2026-10-17 03:11:12 +01:00
• 71c20f73f2 [FlashInfer] Migrate autotune warmup to flashinfer.autotune_v2 managed cache (#55671) 6 people 2026-10-16 19:08:28 -07:00
• b3a7e2ff6e [RL][Frontend] Return weight-version spans on OpenAI completions and chat (#61347) Shuolei Wangandaoshen02 2026-10-17 10:08:07 +08:00
• 6b6a6c5c98 [RL] Support per-parameter source layouts for NCCL M2N (#60934) 3 people 2026-10-16 18:41:17 -07:00
• f8112704d5 [Chore] Use upstream Transformers InklingProcessor (#61883) Harry MellorandClaude Opus 5.5 2026-10-17 02:04:54 +01:00
• f019e3e105 [ROCm] attn res + sigmoid_mul + conv fusions (#52968) 4 people 2026-10-17 03:54:54 +03:00
• 783380af39 [Bugfix] Fix NaN output of the Mamba2 chunk scan kernel on sm90 (#61718) Vadim GimpelsonandClaude 2026-10-17 04:24:51 +04:00
• be25977532 [CI][Tests] Account for FP8 KV-cache rounding in MLA reference (#62067) Wei ZhaoandCodex 2026-10-16 20:10:08 -04:00
• 523d74e2ac [Bugfix][LoRA] Fix garbage/NaN lora_expand output on SM90 with fp32 i… (#61906) 3 people 2026-10-16 18:42:30 -05:00
• a94c315711 [DP] Engines own wave wake-ups; drop the front-end FIRST_REQ (#59559) 3 people 2026-10-17 07:41:54 +08:00
• b544587408 [Core][Feat] Pluggable KVCacheConfigBuilder for platform/model-specific KV cache planning (#53558) 3 people 2026-10-17 07:16:47 +08:00
• 2268d83bd9 [Bugfix][Multimodal] Honor RGB PNG transparency when loading images (#61986) alpandmottopanikeiku 2026-10-16 16:09:12 -07:00
• 28e92eeff2 [ROCm][Perf][Kimi-K3] Use AITER Gluon chunk_kda for KDA prefill (#59552) 4 people 2026-10-17 01:58:56 +03:00
• b088eba944 [Perf][Parser] Stream MiniMax M2 tool argument values incrementally (16 KB call: 5.9 s to 26 ms) (#62035) Afonso RibeiroandFlora Feng 2026-10-16 23:46:02 +01:00
• 596e530675 [Kernel] Add router GEMM kernels for (5120, 256) and BF16 output to ll_bf16_gemm (#62042) 5 people 2026-10-16 15:39:46 -07:00
• 153a17166e [Frontend] Add the Beam reasoning and tool parsers (#62052) 6 people 2026-10-16 14:08:29 -07:00
• 9731f30a03 [Bugfix] Keep VLLM_RAY_BUNDLE_INDICES out of the compile cache key (#61211) Dan BlanaruandClaude Opus 5.5 2026-10-16 22:59:48 +02:00
• 68ebea789a [Perf][MoE] Skip padded work in SM100 FP4 DeepGEMM experts (#59044) 3 people 2026-10-16 16:20:52 -04:00
• cc2e4fecf2 [ROCm] [fusion] support ROCm aiter fusion of per tensor quant + rmsnorm (#52831) 3 people 2026-10-16 15:19:07 -05:00
• 9eb7fb9157 [Build] Remove unused tools/install_deepgemm.sh (#61265) 4 people 2026-10-16 15:57:49 -04:00
• 663e535640 [Perf][Spec Decode] Skip bonus-logit reduction for short drafts in rejection sampling (#61451) cjackal 2026-10-17 04:42:48 +09:00
• 5bf3d9f665 [ROCm][Kimi-K3] Support DSpark adaptive verification on MI355X (#59824) Oxana KorzhandClaude Opus 5.5 2026-10-16 12:53:49 -06:00
• 307c5844ad [ROCm][CI] Enable batch invariance for the ngram scheduling accuracy test (#62048) djramic 2026-10-16 20:36:14 +02:00
• 3dc6def360 [Attention][Spec Decode] Enable varlen Mamba2 decode for adaptive verification (#53036) Guan-Ming Chiuandvllm-agent 2026-10-16 19:23:59 +01:00