Commit Graph - vLLM - iHore Software Repository

iHore/vLLM

Fork 0

ec21f61d82 [Misc] Enable test_fused_moe_wn16 on XPU (#51672) pmanczak 2026-08-10 15:57:02 +02:00
ea3115e30b [CI] Stabilize DP supervisor lifecycle tests (#51557) Taneem IbrahimandWentao Ye 2026-08-10 15:13:52 +02:00
789c4f905e [CI] Bump CUTLASS DSL to 4.6.2 (#51566) Lucas Wilkinson 2026-08-10 09:06:08 -04:00
cf8f3a3bb2 [2/N] Harden Transformers modelling backend multi-modal path (#51657) Harry Mellor 2026-08-10 13:57:12 +01:00
436be94e13 [ROCm][Bugfix] Use TCP store when AITER custom all-reduce is enabled (#51635) vllmellmandtjtanaa 2026-08-10 19:07:14 +08:00
3dafaef027 [Bugfix][Core] Emit --no-{key} for false BooleanOptionalAction flags in YAML config (#51573) Raj Vijay FirkeandHarry Mellor 2026-08-10 16:35:15 +05:30
900d09f91a [Bugfix][Model] Fix Qwen3.5 MTP for text-only checkpoints (#50734) efschu 2026-08-10 13:04:53 +02:00
3a79957b62 [Doc] Add Crusoe Managed Inference deployment guide (#49353) Emmanuel Acheampong 2026-08-10 03:31:34 -07:00
51562de5ab [CI][XPU] Add VLLM_DISABLE_COMPILE_CACHE=1 for other random failed cases in Intel GPU CI (#51604) xiangdong 2026-08-10 17:32:16 +08:00
243c63baf5 [V1][Scheduler] Apply Mamba alignment before encoder caps (#51603)
Jiangyun Zhu Akshat Anand Codex
2026-08-10 16:37:17 +08:00
d89ba6481b [XPU] bump up xpu kernel to v0.1.12.3 (#50441) Kunshang Ji 2026-08-10 16:09:23 +08:00
74c94b9f29 [CI] Upgrade huggingface-hub to 1.27.0 (#51422)
Andreas Karatzas OpenAI Codex Harry Mellor
2026-08-10 02:59:15 -05:00
70b84f0bcb Fix Gemma 4 for upcoming Transformers version (#49797) Harry Mellor 2026-08-10 08:48:26 +01:00
81840a172f [KV Connector] Canonical CPU layout for parallelism-agnostic KV offload (#48414) Itay EtelisandItay Etelis 2026-08-10 10:44:54 +03:00
31cd109f18 [ROCm][CI] Extend ROCm AITER MHA (FA) coverage (#40958) Andreas KaratzasandMicah Williamson 2026-08-10 02:35:13 -05:00
7303c66f68 [Bugfix] Fix lfm2 tool parser dropping calls with brackets or newline… (#48171) Zetian Li - ikunandClaude 2026-08-10 00:10:06 -07:00
61c1dd0966 [EC Connector] Call to EC Connector update_connector_output from scheduler (#49579)
omerpaz95 Nicolò Lucchesimergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
2026-08-10 09:25:57 +03:00
ba1cdcfcf0 [Model][Quantization] Add Ling-3.0-flash-fp8 support (#51265)
zexplorerhj Isotr0pymergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
2026-08-10 14:09:08 +08:00
3b4c86e489 [Bugfix][Quantization] Fix fp32 weight scale for mxfp4 quantization and per-expert checkpoint mapping (#51419) Isotr0py 2026-08-10 14:08:52 +08:00
b22afe45ac [CPU] Enable GPTQ and AWQ quantization for s390x (#51148) Rehan KhanandLi, Jiang <jiang1.li@intel.com> 2026-08-10 09:11:43 +05:30
7ce84b99ce [CPU] Restore linear dispatch for small unquantized GEMMs (#51379) Li, Jiang 2026-08-10 11:29:59 +08:00
a123159f7a Add tiering offloading metrics (#48798) Srinivas Krovvidiandsrinivas_oo7 2026-08-09 21:23:14 -05:00
11ba93f364 [BugFix] Use file:// rendezvous for single-node executors to eliminate startup port races (#50999)
aoshen02 Parsa Mazaheri aoshen02 Claude
2026-08-10 10:00:04 +08:00
0820125ae9 [Bugfix][Parser] Emit REASONING_END for Inkling tool calls that follow no thinking block (#50528) Jason 2026-08-09 21:04:53 -04:00
751f2ccdd3 [Kernel][XPU] Tensor-descriptor operand loads for Triton W8A8 scaled_mm (#47205) Lena OnyshchenkoandKunshang Ji 2026-08-10 02:56:11 +02:00
d6941300fc [BugFix] Scope divergent hybrid cache hits to capable connectors (#50344) Yifan QiaoandOpenAI Codex 2026-08-09 15:38:13 -07:00
83ad767eed [CI] fix docs on main (#51539) Harry Mellor 2026-08-09 10:46:12 +01:00
04d13b5d65 [K3] Allow tpu to import kimi_k3.common (#51529) Jeff (Junze) Ma 2026-08-09 02:06:34 -07:00
c423998642 [KV Offload] Emit self-describing events for partial recurrent blocks (#51243) ChaunceyandOr Ozeri 2026-08-09 16:28:44 +08:00
7f6432cc0d [ROCm][CI] Reuse equivalent ROCm CI images (#48646) Andreas KaratzasandOpenAI Codex 2026-08-09 03:25:03 -05:00
f18e10a7e1 Bump Flashinfer version to 0.6.16.post3 (#50892)
Wei Zhao jiahanc Kevin H. Luu
2026-08-09 03:54:07 -04:00
eb24bc38cf [Bugfix][KV Offload] Handle chunked local attention in offloading scheduler (#51161)
Almog Tavor Or Ozerimergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
2026-08-09 10:25:57 +03:00
1b0ce31f32 [KV Offload] Fix failed-load livelock by marking the lookup verdict as a miss (#49328) Robbie J 2026-08-09 08:25:36 +01:00
f8d03e7741 [Bugfix][Build] Patch stable string memleak fix from 2.14 for 2.13 (#51185) Jane (Yuan) XuandShengqi Chen 2026-08-09 00:12:39 -04:00
073c510c91 [Migration] Migrate bitsandbytes support to OOT plugin (#43529) Isotr0py 2026-08-09 10:27:31 +08:00
dedbf6be8b [V1][Metrics] Preserve prefix-cache stats on zero-output steps (#48668)
Rishi Puri Claude Opus 4.8 Nick Hill
2026-08-08 22:58:38 -03:00
cb1a52aee3 [Build] Upgrade runtime image to Ubuntu 24.04, pick up rdma-core > 44 (#51058)
Tyler Michael Smith OpenAI Codex Michael Goin
2026-08-08 21:17:50 -04:00
7581c56c86 [Test] Add ROCm AITER FP8 MLA prefill accuracy test (#51457) Aarushi JainandAarushi Jain 2026-08-08 19:42:26 -05:00
fbff187d59 [Core] Make the GPU sync check thread-local and fix its suppressors (#51455) Nick Hill 2026-08-08 17:13:40 -07:00
9b0afeb4f6 [Perf] Avoid some more unnecessary GPU<->CPU syncs (#51458) Nick Hill 2026-08-08 17:13:26 -07:00
d608dfabfd [Bugfix][MRV2] Reserve spec-decode lookahead blocks in V2 warmup (#51438)
Nick Hill Rahul Chalamala Janelle Cai
2026-08-08 16:58:43 -07:00
9e6be4a72b [Perf][Sparse MLA] Drop the atomic contention in the index remap (#50365) Nick HillandClaude Opus 5 2026-08-08 16:32:40 -07:00
1c1077c6cc [Bugfix][Parser] Confirm reasoning end when an Inkling content block opens (#49876) Vegetog 2026-08-09 06:07:39 +08:00
75231eff2f [Bugfix][Parser] Prevent Inkling block-end leakage with tools (#51391)
cherry77-cloud Ben Browning Claude Opus 4.8
2026-08-09 00:34:32 +08:00
58d3918e3e [BugFix] Preserve divergent FA hits with external Mamba state (#51468)
Jeff (Junze) Ma Roger Wang Yifan Qiao OpenAI Codex
2026-08-08 05:20:16 -07:00
0fe9a916e8 [Bugfix] Fix LFM2 ShortConv prefix breaking quant ignore list (#51495) ylsunandClaude Opus 5 2026-08-08 19:27:13 +08:00
653ebb52df Add torch compile for qwen3_vl encoder (#40116) Tianyu Guo 2026-08-08 16:23:18 +08:00
643c125fab [ROCm][CI] Baseline legacy extensions in the Torch ABI audit (#50805) Andreas Karatzas 2026-08-08 01:01:04 -05:00
6e18901195 [Bugfix][MM] Avoid device sync in FusedInputNorm initialization (#51435) Andreas Karatzas 2026-08-08 00:32:23 -05:00
44351f81d5 [CI] Refresh hybrid Model Runner V2 coverage (#51410) Jiangyun Zhu 2026-08-08 13:10:54 +08:00
700d39b558 [CPU] Optimize routed FP8/MXFP4 MoE GEMM dispatch (#50949) Tianmu Li 2026-08-07 21:56:14 -07:00
d3621c1eb3 [Bugfix][Multimodal] Fix PyNvVideoCodec video backend returning NCHW instead of NHWC (#51076) Junpu Yu 2026-08-07 21:49:06 -07:00
22a175921c [BugFix] Reject invalid data-parallel RPC ports (#51469) aoshen02andOpenAI Codex 2026-08-08 12:23:26 +08:00
d21fed9522 [Bugfix][models_multimodal] Remote HF python code misses importing class (#51427) Gyula ChinoradszkiandCursor Agent 2026-08-08 06:09:26 +02:00
a828536fb5 [Bugfix][multi_modal] Fix pos_ids being unitialized for minicpmv2.6 in hf runner (#51432) music-dino 2026-08-08 06:09:12 +02:00
12da9b23ae [CI] Guard remote-code Transformers compatibility (#51451) Andreas Karatzas 2026-08-07 23:05:49 -05:00
5eaa70dac5 [Bugfix][Quantization] Fix INT8 W8A8 MoE crash in TritonExperts (#51411) djramic 2026-08-08 05:06:58 +02:00
f7ef489e93 [Bugfix] Fix ZMQ port TOCTOU race in shm_broadcast MessageQueue (#50960)
aoshen02 RTCartist aoshen02 Claude Fable 5
2026-08-08 09:19:02 +08:00
7f58e8294a [Kimi][MM] disable kimi_vit's dynamic torch.compile for TPU (#51196) Linkun 2026-08-07 17:42:51 -07:00
1c008a346c [Bugfix] Close usage telemetry HTTP sessions (#51219) Nils Matteson 2026-08-07 17:29:36 -07:00
e644c8cd8c [Perf] Optimize DeepSeek V3.2 sequence parallelism (#51434) Woosuk Kwon 2026-08-07 16:51:01 -07:00
c39076feff [CI Test] Add specific unit test for mrv2 offloading (#51440) Wentao Ye 2026-08-07 18:49:45 -04:00
a801e71cb8 [Perf] Improve --linear-backend filtering (#48735) Andrii SkliarandAleksandr Skliar 2026-08-08 00:15:54 +02:00
46b5864054 [Perf] Narrow DeepSeek V3.2 eager CUDA graph region (#51425) Woosuk Kwon 2026-08-07 14:37:43 -07:00
c8fe1d5715 [Spec Decode] Register Qwen3.6 dSpark acceptance coverage (#51310) Michael GoinandOpenAI Codex 2026-08-07 17:07:55 -04:00
9823714346 [Bugfix] Drop stale layer kwarg from online MXFP4 kernel creation (fix precommit) (#51442) Nick Hill 2026-08-07 13:49:00 -07:00
27d7303802 [CI] Fix Batch Invariance (B200) (#51417) Jiangyun Zhu 2026-08-08 04:45:29 +08:00
4a11ed5bc4 [DSv32/GLM Perf] Skip short prefill topk for dense mha layer, 97.9% kernel level latency reduction (#51298) Wentao Ye 2026-08-07 16:30:42 -04:00
99950b0b86 [Profiler] Stamp vLLM version/commit into torch profiler trace metadata (#51389) Elvir Crnčević 2026-08-07 22:29:59 +02:00
1c94e8dc7d Add NVFP4 KV 4-over-6 scale search (#45187) Wei-Ming Chen 2026-08-07 13:22:27 -07:00
70456e5e6f [Bugfix] when loading weights skip empty expert bias if model does not support them (#50937) Walter Beller-MoralesandHarry Mellor 2026-08-07 15:45:52 -04:00
ac70ce96e0 [Frontend] Disable uvicorn signal handlers instead of racing them (#50916) Nick Hill 2026-08-07 12:40:24 -07:00
3518110f2f [Online quantization] Add online MXFP4 quantization support (#49347) fxmarty-amd 2026-08-07 21:39:46 +02:00
a671679e9f [MRv2 Feature] MR v2 weight offloading support (#51413) Wentao Ye 2026-08-07 15:20:21 -04:00
021b7d985b [Perf] Raise Blackwell CUDA graph capture default to 1024 (#49390) Lucas Wilkinson 2026-08-07 15:18:14 -04:00
56a4b63d44 [K3 Perf] Optimize k3 dspark fused kv, 4.5~4.6x kernel performance improvement (#50585) Wentao Yeandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com> 2026-08-07 15:16:11 -04:00
0df620d429 [Test] Add packed DeepSeek-V4 KV zeroer geometry regression (#51288) coltonottleyandColton Ottley 2026-08-07 13:11:09 -06:00
a0056e103e [Test] Add ROCm AITER MLA op registration and env gating tests (#50930) Aarushi JainandAarushi Jain 2026-08-07 14:07:31 -05:00
ebb2972562 [ROCm][CI][Bugfix] Do not microbatch a step that splits a prefix from its writer (#51402) stefankoncarevic 2026-08-07 20:55:34 +02:00
4eccf906ca [Attention] Mamba attention module refactor - Final part (#44857) wangxiyuan 2026-08-08 02:27:10 +08:00
45273b8dcb [1/N] Harden Transformers modelling backend multi-modal path (#51408) Harry Mellor 2026-08-07 18:30:43 +01:00
fcde8e1460 [Refactor] refactor humming linear and moe backends to use explicit layer configs (#49610) Jinzhen LinandOpenAI Codex 2026-08-08 01:03:19 +08:00
e229fdbc87 [ROCm] Add tuned selective_state_update float32 config for AMD Instinct MI325X (#50007) vanshbhatia-amd 2026-08-07 12:58:36 -04:00
34c1cd20a5 [Bugfix][ROCm] Fix ROCM_AITER_FA & ROCM_AITER_UNIFIED_ATTN QK-Norm+RoPE+KVCache fusion for the packed KV-cache [BLOCKS, HEADS, BLOCK_SIZE, 2*HEAD_DIM] layout (#49373) Jack HuandAndreas Karatzas 2026-08-07 12:54:55 -04:00
7e85d3a42c [ROCm] Enable pinned memory on supported WSL2 kernels (#50126) Flora Cui 2026-08-07 23:43:42 +08:00
448344c0e2 [Bugfix] Fix get_open_port() livelock on DP-reserved ports and cover get_open_ports_list (#50965)
aoshen02 Chenglun Hu ferkans-amir aoshen02 Claude Fable 5
2026-08-07 22:39:33 +08:00
47228db84c [Bugfix][KV-transfer] MoRIIO: per-layer READ-completion barrier in wait_for_layer_load (#48534)
limeward Jaeyoun Kim Claude Opus 4.8 Redwan Khan harishk-mangoboost rkhan055 Peiran Qin QinPR
2026-08-07 07:23:33 -07:00
d4ecb75ba2 [PD][NixlPush][Bugfix] Fix prefix caching (#48758) Nicolò Lucchesi 2026-08-07 16:21:17 +02:00
f2bfad9167 [Model] Enable Qwen3.8 for AMD Rocm (#50068) haic0 2026-08-07 21:19:07 +08:00
a231c5ceac [Bugfix] Skip fetching revision for model when model and weights_model are different (#51260) music-dinoandHarry Mellor 2026-08-07 14:55:50 +02:00
d5aae2b464 Fix ROCm architecture import on non-ROCm platforms (#51357) Xiaochang WuandKunshang Ji 2026-08-07 20:32:56 +08:00
ae934ba8a5 feat: extended EPLB support for Mistral Large 3 and additional MoE backends (#48355) Julien Debache 2026-08-07 14:32:52 +02:00
8d9b52f7c2 [XPU] quick fix online quantization UT break (#51365) Yan Ma 2026-08-07 18:25:31 +08:00
6b5bec7bed [Misc] Add and enable Triton kernel unit tests on XPU (#45694) pmanczakandpmanczak 2026-08-07 11:53:03 +02:00
5ec47f3e48 [PD][PushConnector] Record last activity of remotes to allow clean up of stale ones (#50234) Nicolò Lucchesi 2026-08-07 11:43:29 +02:00
4f76c8ad9d [Bugfix][Platform] Stop re-initializing NVML on every device-capability check (fixes #50381) (#50393) Sebastian WooandClaude 2026-08-07 01:30:00 -07:00
b8db7f4abd [Bugfix][Quantization] Fix dynamic INT8 W8A8 MoE config being built as W8A16 (#50833) Hank_andCyrus Leung 2026-08-07 16:21:13 +08:00
c84789c40b [Refactor] Remove kernel dead code (#51051) Wentao Ye 2026-08-07 04:20:19 -04:00
da788334bc Support DeepSeek-V4 AMD Quark NVFP4 with emulation kernel (#47972)
jimmy-adams Andreas Karatzas huji Cursor
2026-08-07 16:19:31 +08:00
0de0362ea1 [ROCm][CI] Loosen block-FP8 fused MoE test tolerance for large-K shapes (#48847)
stefankoncarevic Andreas Karatzasmergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
2026-08-07 10:19:26 +02:00

... 35 36 37 38 39 ...

Commit Graph Select branches Hide Pull Requests main patches sm86 Mono Color

Commit Graph

Select branches

Hide Pull Requests

main

patches

sm86