sglang

deepseek_common

[Feature] Unified memory: support decode context parallelism for Kimi-Linear (#36890 )

2026-09-01 12:44:26 -07:00

dots3_common

[Kernel] Replace dsv3_router_gemm with the unified tiny GEMM (#34693 )

2026-09-01 23:00:21 +08:00

inkling_common

[Config] Round 5.1: the published-side readers ask the bags, and a platform fact gets one address (#37086 )

2026-08-30 02:18:33 -07:00

afmoe.py

[NPU] Acc fix for afmoe model introduced by topk refactor. (#31280 )

2026-07-27 15:40:49 +08:00

apertus.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

arcee.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

baichuan.py

[Fix] model init / XPU / transformers-v5 / bench-image fixes (#28292 )

2026-06-23 12:58:13 +08:00

bailing_moe_linear.py

[Config] Round 5.1: the published-side readers ask the bags, and a platform fact gets one address (#37086 )

2026-08-30 02:18:33 -07:00

bailing_moe_nextn.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

bailing_moe_v3.py

[Config] Round 5.1: the published-side readers ask the bags, and a platform fact gets one address (#37086 )

2026-08-30 02:18:33 -07:00

bailing_moe.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

bert.py

config: read resolved config via namespace accessors (#33013 )

2026-07-31 15:06:59 -07:00

chatglm.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

clip.py

[diffusion] chore: reuse SRT CLIP encoder blocks (#35004 )

2026-08-17 19:51:51 +08:00

cohere2_moe.py

[Cohere Command-A-Plus] Optimize decode and BCG capture on SM10X (#36624 )

2026-08-31 19:37:47 -07:00

cohere2_vision.py

model: support Command A plus (#26106 )

2026-06-03 11:23:04 +08:00

commandr.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

dbrx.py

[MoE Refactor] [NPU] Refactor Ascend MoE implementation to reduce code duplication and align with community design (#25663 )

2026-07-15 14:59:42 +03:00

deepseek_janus_pro.py

…

deepseek_nextn.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

deepseek_ocr.py

moe: the shared-experts-fusion decision is a per-runner value the loader installs (#33889 )

2026-08-07 22:42:58 -07:00

deepseek_v2.py

[Kernel] Replace dsv3_router_gemm with the unified tiny GEMM (#34693 )

2026-09-01 23:00:21 +08:00

deepseek_v4_dspark.py

[Config] Round 5.1: the published-side readers ask the bags, and a platform fact gets one address (#37086 )

2026-08-30 02:18:33 -07:00

deepseek_v4_nextn.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

deepseek_v4.py

[SM120] DeepSeek-V4: DeepGEMM paged-MQA indexer +FP4 MoE+ page-split (#29927 )

2026-09-02 15:49:59 -07:00

deepseek_vl2.py

moe: the shared-experts-fusion decision is a per-runner value the loader installs (#33889 )

2026-08-07 22:42:58 -07:00

deepseek.py

[MoE Refactor] [NPU] Refactor Ascend MoE implementation to reduce code duplication and align with community design (#25663 )

2026-07-15 14:59:42 +03:00

dflash.py

[Model] Support Nemotron 3.5 Lightning speculative decoding (#36186 )

2026-08-25 16:43:58 -07:00

dots3_nextn.py

[Model] Complete dots.note.omni support with native encoders, video preprocessing, and MTP decoding (#33829 )

2026-08-22 14:19:14 +08:00

dots3.py

[Model] Complete dots.note.omni support with native encoders, video preprocessing, and MTP decoding (#33829 )

2026-08-22 14:19:14 +08:00

dots_ocr.py

…

dots_vlm_vit.py

vlm: batch cross-request vit encoding and reuse attention metadata (#24013 )

2026-07-18 18:50:33 +08:00

dots_vlm.py

moe: the shared-experts-fusion decision is a per-runner value the loader installs (#33889 )

2026-08-07 22:42:58 -07:00

dspark.py

[Spec] Add LFM2 and LFM2-MoE DSpark speculative decoding support (#31041 )

2026-08-30 20:03:59 -07:00

ernie4_eagle.py

…

ernie4.py

[CPU] Enable ERNIE models on CPU (#35222 )

2026-08-27 10:13:10 +08:00

ernie45_moe_vl.py

Fix rope config compatibility and VL/transformers-fallback weight loading (#31575 )

2026-08-17 14:53:59 +08:00

ernie45_vl.py

vlm: batch cross-request vit encoding and reuse attention metadata (#24013 )

2026-07-18 18:50:33 +08:00

exaone4.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

exaone_moe_mtp.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

exaone_moe.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

exaone.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

falcon_h1.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

gemma2_reward.py

[Score API] Add return_pooled_hidden_states to Scoring API for SequenceClassification / RewardModel (#22427 )

2026-04-15 14:58:56 -07:00

gemma2.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

gemma3_causal.py

optimize: optimize EmbeddingGemma prefill performance (#32383 )

2026-07-27 17:34:29 +08:00

gemma3_mm.py

feat(model_runner): remove pool/backend refs from ForwardBatch via ForwardContext (#25983 )

2026-05-21 14:01:49 -07:00

gemma3n_audio.py

…

gemma3n_causal.py

[Fix] model init / XPU / transformers-v5 / bench-image fixes (#28292 )

2026-06-23 12:58:13 +08:00

gemma3n_mm.py

[Fix] model init / XPU / transformers-v5 / bench-image fixes (#28292 )

2026-06-23 12:58:13 +08:00

gemma4_audio.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

gemma4_causal.py

[CPU] Add support for Gemma4 on Xeon (#22498 )

2026-08-17 10:52:26 +08:00

gemma4_mm.py

[CPU] Add support for Gemma4 on Xeon (#22498 )

2026-08-17 10:52:26 +08:00

gemma4_mtp.py

fix(gemma4): quantize MTP bridge projections (#32440 )

2026-08-19 12:41:52 -07:00

gemma4_unified.py

[model] support encoder-free unified Text/Vision/Audio model (#27167 )

2026-06-03 23:58:06 +08:00

gemma4_vision.py

[Config] Round 5.1: the published-side readers ask the bags, and a platform fact gets one address (#37086 )

2026-08-30 02:18:33 -07:00

gemma.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

glm4_moe_lite_nextn.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

glm4_moe_lite.py

Fix stale GLM MoE routing after runtime weight updates (#35883 )

2026-08-30 14:13:34 -07:00

glm4_moe_nextn.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

glm4_moe.py

Fix stale GLM MoE routing after runtime weight updates (#35883 )

2026-08-30 14:13:34 -07:00

glm4.py

[refactor] Move MLP collective flags onto ForwardFlags (#30802 )

2026-07-10 17:43:39 -07:00

glm4v_moe.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

glm4v.py

config: read resolved config via namespace accessors (#33013 )

2026-07-31 15:06:59 -07:00

glm_image_vl.py

config: read resolved config via namespace accessors (#33013 )

2026-07-31 15:06:59 -07:00

glm_ocr_nextn.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

glm_ocr.py

config: read resolved config via namespace accessors (#33013 )

2026-07-31 15:06:59 -07:00

glmasr.py

fix: fix transcription & audio-understanding for ASR/audio/speech models (#32611 )

2026-08-19 18:44:23 +08:00

gpt2.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

gpt_bigcode.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

gpt_j.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

gpt_oss.py

[Config] Round 5.1: the published-side readers ask the bags, and a platform fact gets one address (#37086 )

2026-08-30 02:18:33 -07:00

granite.py

[Fix] Read the granite sinks dtype from the exec bag, not the legacy global shim (#35921 )

2026-08-21 18:13:18 -07:00

granitemoe.py

Add SGLang Granite SWA support via existing Granite models (#35794 )

2026-08-21 11:13:15 -07:00

granitemoehybrid.py

Add SGLang Granite SWA support via existing Granite models (#35794 )

2026-08-21 11:13:15 -07:00

grok.py

[Kernel] Classification cleanup: unify _jit_ naming, drop empty/model groups, add elementwise (RFC #29630 ) (#32148 )

2026-07-23 13:47:02 +08:00

hrm_text.py

[refactor] Adopt get_parallel() everywhere and close out the parallel wrapper surface (#30492 )

2026-07-09 02:09:39 -07:00

hunyuan_v3_nextn.py

[Fix] Load HunyuanV3 NextN final_layernorm into the draft head's output norm (#30331 )

2026-07-13 20:37:36 +08:00

hunyuan_v3.py

[Attention Backend] Add HPC-Ops attention backend (#30540 )

2026-07-22 22:06:22 +08:00

hunyuan.py

chore: cleanup garbage code (#29770 )

2026-07-02 16:14:01 +08:00

idefics2.py

vlm: batch cross-request vit encoding and reuse attention metadata (#24013 )

2026-07-18 18:50:33 +08:00

inkling.py

config: retire the hidden global fallbacks and the mamba-extra-buffer instance reads (#34080 )

2026-08-09 14:43:28 -07:00

internlm2_reward.py

[Score API] Add return_pooled_hidden_states to Scoring API for SequenceClassification / RewardModel (#22427 )

2026-04-15 14:58:56 -07:00

internlm2.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

interns1.py

[Bugfix] Fix Intern-S1 FP8 expert count lookup (#28629 )

2026-06-18 17:49:59 +08:00

interns1pro.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

interns2_mobius.py

Support Intern-S2-Mobius FP8 (#34908 )

2026-08-19 10:58:01 -07:00

interns2preview.py

Support Intern-S2-Preview (#24875 )

2026-05-10 22:17:30 +08:00

internvl.py

fix(xpu): enable compressed-tensors FP8 W8A8 on XPU (RedHatAI FP8-dynamic models) (#33057 )

2026-08-24 12:53:48 +08:00

iquest_loopcoder.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

jet_nemotron.py

[Kernel] Relocate vendored fla and mamba kernel trees to sglang.kernels (RFC #29630 , Phase 2.5, 7/7) (#30795 )

2026-07-15 12:52:15 +08:00

jet_vlm.py

…

kimi_k3_vl.py

config: retire the alias-form process-global config reads

2026-08-05 19:29:20 -07:00

kimi_k3.py

[AMD] Perf Kimi-K3 fuse ROCm KDA decode boundary (#34198 )

2026-09-01 18:35:48 -07:00

kimi_k25_eagle3.py

Fixing MXFP8 online quantization pipeline (#31510 )

2026-07-28 21:26:13 -07:00

kimi_k25.py

config: a parallel size has one spelling; a patched scope declares its own (#36621 )

2026-08-27 12:56:42 -07:00

kimi_linear.py

[Model] Support Ling-3.0-flash (BailingMoeV3) (#33561 )

2026-08-26 17:27:23 -07:00

kimi_vl_moonvit.py

[Kimi] Support kimi-k3 (#32541 )

2026-08-04 13:22:49 -07:00

kimi_vl.py

moe: the shared-experts-fusion decision is a per-runner value the loader installs (#33889 )

2026-08-07 22:42:58 -07:00

laguna.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

lfm2_dspark.py

[Spec] Add LFM2 and LFM2-MoE DSpark speculative decoding support (#31041 )

2026-08-30 20:03:59 -07:00

lfm2_moe.py

[Spec] Add LFM2 and LFM2-MoE DSpark speculative decoding support (#31041 )

2026-08-30 20:03:59 -07:00

lfm2_vl.py

…

lfm2.py

[Spec] Add LFM2 and LFM2-MoE DSpark speculative decoding support (#31041 )

2026-08-30 20:03:59 -07:00

lightonocr.py

[Fix] model init / XPU / transformers-v5 / bench-image fixes (#28292 )

2026-06-23 12:58:13 +08:00

llada2.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

llama4.py

[refactor] Move MLP collective flags onto ForwardFlags (#30802 )

2026-07-10 17:43:39 -07:00

llama_classification.py

Fix reward/classification models broken by load_weights v2 dispatch (#28671 ) (#31988 )

2026-07-21 18:47:49 -07:00

llama_eagle3.py

config: read resolved config via namespace accessors (#33013 )

2026-07-31 15:06:59 -07:00

llama_eagle.py

runtime: Add flashinfer rmsnorm + quant fusion support SM90, SM100, SM120- #32994 (#33471 )

2026-08-09 20:15:03 +08:00

llama_embedding.py

…

llama_reward.py

Fix reward/classification models broken by load_weights v2 dispatch (#28671 ) (#31988 )

2026-07-21 18:47:49 -07:00

llama.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

llava.py

perf: migrate Req token-id storage to array.array('q') in Scheduler (#25098 )

2026-05-22 10:51:07 -07:00

llavavid.py

perf: migrate Req token-id storage to array.array('q') in Scheduler (#25098 )

2026-05-22 10:51:07 -07:00

locate_anything.py

model: support nvidia/LocateAnything-3B (#28958 )

2026-06-30 00:16:42 +08:00

longcat_flash_nextn.py

ci: add LongCat-Flash-Lite-FP8 8-GPU nightly test + fix NextN rope_theta (#32125 )

2026-07-24 19:45:28 +08:00

longcat_flash.py

[Config] Round 5.1: the published-side readers ask the bags, and a platform fact gets one address (#37086 )

2026-08-30 02:18:33 -07:00

mellum.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

midashenglm.py

…

mimo_audio.py

[Config] Round 5.1: the published-side readers ask the bags, and a platform fact gets one address (#37086 )

2026-08-30 02:18:33 -07:00

mimo_mtp.py

Support speculative decoding on CPU (#27862 )

2026-07-09 10:27:09 +08:00

mimo_v2_asr.py

Fix MiMo-V2 on Blackwell: FA3 fallback and TP-aware audio weight loading (#31343 )

2026-07-15 13:26:27 -07:00

mimo_v2_nextn.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

mimo_v2.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

mimo_vl.py

config: read resolved config via namespace accessors (#33013 )

2026-07-31 15:06:59 -07:00

mimo.py

…

mindspore.py

[Spec] Converge DP-attention spec width scaling onto num_tokens_per_req (#31244 )

2026-07-15 17:57:08 -07:00

minicpm3.py

Delete sgl-kernel AOT bmm_fp8, use flashinfer.bmm_fp8 (#31202 )

2026-07-22 07:44:47 +08:00

minicpm.py

[Feature] Add MiniCPM-SALA support (#30360 )

2026-08-24 02:25:16 -07:00

minicpmo.py

[Deps] Bump transformers to 5.12.1 (#29393 )

2026-06-30 00:54:01 -07:00

minicpmv_vit.py

vlm: batch cross-request vit encoding and reuse attention metadata (#24013 )

2026-07-18 18:50:33 +08:00

minicpmv.py

moe: the shared-experts-fusion decision is a per-runner value the loader installs (#33889 )

2026-08-07 22:42:58 -07:00

minimax_m2.py

[XPU] Lazily import tvm_ffi-dependent all_reduce kernel in minimax_m2 (#35290 )

2026-08-28 02:18:03 -07:00

minimax_m3_vl.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

minimax_m3.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

minimax_vl_common.py

config: read resolved config via namespace accessors (#33013 )

2026-07-31 15:06:59 -07:00

ministral3.py

[Bugfix] Fix Ministral3 init argument forwarding (#29111 )

2026-06-25 18:26:45 +08:00

mistral_eagle.py

debug followup (#24058 )

2026-04-29 23:03:27 +08:00

mistral_large_3_eagle.py

Fix Mistral-Large-3 EAGLE draft skipping DeepseekV2Model.__init__ (#33785 )

2026-08-06 13:59:28 -07:00

mistral_large_3.py

chore: add vLLM SPDX copyright headers to ported files (#25182 )

2026-05-13 15:17:30 -07:00

mistral.py

Bugfix (#24027 )

2026-04-29 21:13:51 +08:00

mixtral_quant.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

mixtral.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

mllama4.py

config: read resolved config via namespace accessors (#33013 )

2026-07-31 15:06:59 -07:00

mllama.py

[CPU] Padding for dim divisibility in TP3/6 cases (#20072 )

2026-07-02 13:14:39 +08:00

moss_vl.py

config: read resolved config via namespace accessors (#33013 )

2026-07-31 15:06:59 -07:00

muse_glimmer.py

[Feature] Add Muse Glimmer model support (#34262 )

2026-08-11 15:41:52 -07:00

nano_nemotron_vl.py

feat: make mm_inputs msgpack-native (#29656 )

2026-08-20 14:30:07 -07:00

nemotron_h_mtp.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

nemotron_h_utils.py

Support Flashinfer one-sided A2A + CuteDSL MoE for Nemotron Ultra (#28309 )

2026-07-15 18:14:20 -07:00

nemotron_h.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

nemotron_nas.py

chore: add vLLM SPDX copyright headers to ported files (#25182 )

2026-05-13 15:17:30 -07:00

nvila_lite.py

[Cherry pick to release/v0.5.15] Fix NVILA weight loading (#30400 )

2026-07-07 13:21:55 -07:00

nvila.py

Fix nightly CI: NVFP4 cuda-graph crash, NVILA batching, CuTe paged-KV zero-size, Kimi-VL OOM (#32118 )

2026-07-29 01:39:55 -07:00

olmo2.py

Fix rope config compatibility and VL/transformers-fallback weight loading (#31575 )

2026-08-17 14:53:59 +08:00

olmo.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

olmoe.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

opt.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

orion.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

paddleocr_vl.py

[perf] overlap page preprocessing, pack the vit, enable prefill CUDA graph for paddle-ocr (#35318 )

2026-08-19 08:20:55 +08:00

parakeet.py

chore: add vLLM SPDX copyright headers to ported files (#25182 )

2026-05-13 15:17:30 -07:00

persimmon.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

phi3_small.py

[Fix] model init / XPU / transformers-v5 / bench-image fixes (#28292 )

2026-06-23 12:58:13 +08:00

phi4mm_audio.py

…

phi4mm_utils.py

…

phi4mm.py

chore: add vLLM SPDX copyright headers to ported files (#25182 )

2026-05-13 15:17:30 -07:00

phi.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

phimoe.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

pixtral.py

moe: the shared-experts-fusion decision is a per-runner value the loader installs (#33889 )

2026-08-07 22:42:58 -07:00

points_v15_chat.py

…

qwen2_5_vl.py

[vlm] fix: preserve per-request vit graph metadata for qwen-vl (#37043 )

2026-08-30 21:02:14 +08:00

qwen2_audio.py

…

qwen2_classification.py

Fix reward/classification models broken by load_weights v2 dispatch (#28671 ) (#31988 )

2026-07-21 18:47:49 -07:00

qwen2_eagle.py

runtime: Add flashinfer rmsnorm + quant fusion support SM90, SM100, SM120- #32994 (#33471 )

2026-08-09 20:15:03 +08:00

qwen2_moe.py

[Config] Round 5.1: the published-side readers ask the bags, and a platform fact gets one address (#37086 )

2026-08-30 02:18:33 -07:00

qwen2_rm.py

Fix reward/classification models broken by load_weights v2 dispatch (#28671 ) (#31988 )

2026-07-21 18:47:49 -07:00

qwen2_vl.py

vlm: batch cross-request vit encoding and reuse attention metadata (#24013 )

2026-07-18 18:50:33 +08:00

qwen2.py

runtime: Add flashinfer rmsnorm + quant fusion support SM90, SM100, SM120- #32994 (#33471 )

2026-08-09 20:15:03 +08:00

qwen3_5_mtp.py

[Bugfix] Load Qwen3.5 MTP embedding under PP (#37471 )

2026-09-02 15:19:40 -07:00

qwen3_5_text.py

qwen 3.8 rebase (#35758 )

2026-08-28 20:41:34 -07:00

qwen3_5.py

[FlashInfer v0.6.18] add FlashInfer CuTe DSL NVFP4 W4A16 mode (#35120 )

2026-08-31 18:47:30 -07:00

qwen3_asr.py

…

qwen3_classification.py

[Model] Qwen3ForPooledOutput: forward get_input_embeddings to inner model (#23434 )

2026-04-29 12:25:06 -07:00

qwen3_embedding.py

model: serve bare Qwen3Model backbone natively as an embedding model (#32457 )

2026-07-27 15:49:58 +08:00

qwen3_moe_mtp.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

qwen3_moe.py

📝 [NPU] Use vendor-neutral wording in quantization comments (#36768 )

2026-08-29 23:05:35 +03:00

qwen3_next_mtp.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

qwen3_next.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

qwen3_omni_moe.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

qwen3_rm.py

…

qwen3_vl_moe.py

[BugFix][VLM] keep Qwen3-VL MoE inference deepstack order (#34690 )

2026-08-28 17:02:16 +08:00

qwen3_vl.py

[vlm] fix: preserve per-request vit graph metadata for qwen-vl (#37043 )

2026-08-30 21:02:14 +08:00

qwen3.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

qwen.py

Fix rope config compatibility and VL/transformers-fallback weight loading (#31575 )

2026-08-17 14:53:59 +08:00

radio.py

chore: add vLLM SPDX copyright headers to ported files (#25182 )

2026-05-13 15:17:30 -07:00

registry.py

chore: add vLLM SPDX copyright headers to ported files (#25182 )

2026-05-13 15:17:30 -07:00

roberta.py

…

sarashina2_vision.py

[Rust Server] Add e2e latency metadata and fix Sarashina import (#35125 )

2026-08-18 02:08:07 -07:00

sarvam_moe.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

sdar_moe.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

sdar.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

siglip2.py

vlm: batch cross-request vit encoding and reuse attention metadata (#24013 )

2026-07-18 18:50:33 +08:00

siglip.py

[diffusion] chore: reuse SRT SigLIP in Pi0.5 (#34992 )

2026-08-17 19:33:36 +08:00

solar.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

spark2_5.py

Rename Spark3 to Spark2.5 (#36416 )

2026-08-26 00:04:46 -07:00

stablelm.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

starcoder2.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

step3_vl_10b.py

[FIX]Fix Step3-VL multi-image embedding and local patch splitting (#24970 )

2026-06-17 10:31:32 -07:00

step3_vl.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

step3p5_mtp.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

step3p5.py

config: a parallel leaf with no live counterpart is read bare (#36620 )

2026-08-27 12:56:11 -07:00

step3p7.py

model: support Step-3.7-Flash (#26565 )

2026-05-29 08:00:54 +08:00

teleflm.py

…

torch_native_llama.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

transformers.py

[Fix] Transformers-fallback (GPT-NeoX) + KV pool config (DeepSeek-VL2) (#35244 )

2026-08-31 11:05:50 +08:00

unlimited_ocr.py

model: support baidu unlimited-ocr (#29186 )

2026-06-27 23:36:19 +08:00

utils.py

[AMD][Fix] Qwen3.5: guard zero-grid launch in fused_qk_gemma_rmsnorm(_with_gate) (HIP invalid configuration on idle DP rank) (#31794 )

2026-08-15 21:02:54 -07:00

voxtral.py

chore: add vLLM SPDX copyright headers to ported files (#25182 )

2026-05-13 15:17:30 -07:00

whisper.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

xverse_moe.py

[MoE Refactor] [NPU] Refactor Ascend MoE implementation to reduce code duplication and align with community design (#25663 )

2026-07-15 14:59:42 +03:00

xverse.py

Add get_parallel(): a structured accessor for parallel-topology state (#28567 )

2026-06-17 20:23:43 -07:00

yivl.py

Upgrade transformers from 5.5.4 to 5.6.0 (#23525 )

2026-04-26 22:33:54 -07:00

zaya.py

feat(short-conv): shared ShortConvAttnBackend for ZAYA1 CCA + LFM2 short conv (#29867 )

2026-07-01 20:08:00 -07:00