sglang

deepseek_common

Fix OOM in DeepSeek weight loading by deferring dict(weights) materialization (#17744 )

2026-01-31 13:59:00 +08:00

afmoe.py

Add AFMoE model implementation (#13216 )

2026-01-16 20:35:42 +08:00

apertus.py

Depreate global_server_args_dict (#11528 )

2025-10-13 19:34:43 +08:00

arcee.py

Depreate global_server_args_dict (#11528 )

2025-10-13 19:34:43 +08:00

baichuan.py

accuracy enhancement for baichuan2-13B for npu (#16868 )

2026-01-26 16:14:35 +08:00

bailing_moe_nextn.py

Fix the scenario where eh_proj is quantized in the bailing moe nextn weights (#17808 )

2026-01-29 21:11:47 -08:00

bailing_moe.py

feat: Add Ling Flash v2.0 support for Eagle3 (#15119 )

2026-01-31 23:57:45 -08:00

bert.py

[NPU] bugfix with Kimi-k2 and bge-reranker-v2 model (#17478 )

2026-01-22 22:02:05 +08:00

chatglm.py

…

clip.py

refactor: cleanup vision attention related codes (#13228 )

2025-11-16 22:44:40 +08:00

commandr.py

[new-model] Add support for Cohere2ForCausalLM behind Command-A and Command-R Models (#16927 )

2026-01-21 12:28:33 -08:00

dbrx.py

[6/N] MoE Refactor: Cleanup MoE-related configs (#8849 )

2025-08-14 21:14:53 -07:00

deepseek_janus_pro.py

refactor: cleanup vision attention related codes (#13228 )

2025-11-16 22:44:40 +08:00

deepseek_nextn.py

Use attn_tp_group for all reduce in token embedding (#17403 )

2026-01-20 23:56:21 +08:00

deepseek_ocr.py

model: support DeepSeek-OCR-2 (#17897 )

2026-01-30 09:49:51 +08:00

deepseek_v2.py

[Tiny] Fix grammar in shared experts fusion log messages (#18043 )

2026-01-31 13:14:25 -08:00

deepseek_vl2.py

[Refactor] simplify multimodal data processing (#8107 )

2025-07-20 21:43:09 -07:00

deepseek.py

[6/N] MoE Refactor: Cleanup MoE-related configs (#8849 )

2025-08-14 21:14:53 -07:00

dots_ocr.py

[Lint] Add python/sglang to ruff F401 checks and remove unused imports in files (#11685 )

2025-10-17 16:49:46 -07:00

dots_vlm_vit.py

Convert cu_seqlens to CPU for npu_flash_attention_unpad operator (#15434 )

2026-01-04 08:16:42 +08:00

dots_vlm.py

feat: support EPD disaggregation (#12263 )

2025-12-14 22:30:08 +08:00

ernie4_eagle.py

Add ernie4.py for ERNIE-4.5 (#7657 )

2025-08-08 00:55:48 -07:00

ernie4.py

[Feature] Enable return routed experts (#12162 )

2025-12-21 15:16:43 +08:00

ernie45_moe_vl.py

[Model] Add Ernie4.5 VL model support (#15679 )

2026-01-25 22:36:29 -08:00

ernie45_vl.py

[Model] Add Ernie4.5 VL model support (#15679 )

2026-01-25 22:36:29 -08:00

exaone4.py

[Model] Add support for EXAONE-4.0 Model (#8205 )

2026-01-27 14:08:24 +08:00

exaone_moe_mtp.py

[Model] Add K-EXAONE model support (#16294 )

2026-01-30 20:01:14 +08:00

exaone_moe.py

[Model] Add K-EXAONE model support (#16294 )

2026-01-30 20:01:14 +08:00

exaone.py

…

falcon_h1.py

Use attn tp group in embedding for more models (#17570 )

2026-01-24 13:37:44 +08:00

gemma2_reward.py

…

gemma2.py

Fix incorrect default get_hidden_dim logic (#8987 )

2025-08-09 00:25:38 -07:00

gemma3_causal.py

vlm: refactor engine vlm params and support processor output as input (#14091 )

2025-12-20 18:31:24 +08:00

gemma3_mm.py

vlm: refactor engine vlm params and support processor output as input (#14091 )

2025-12-20 18:31:24 +08:00

gemma3n_audio.py

…

gemma3n_causal.py

…

gemma3n_mm.py

[Lint] Add python/sglang to ruff F401 checks and remove unused imports in files (#11685 )

2025-10-17 16:49:46 -07:00

gemma.py

[Feature] Layer-wise Prefill (#7634 )

2025-07-17 01:57:46 +08:00

glm4_moe_lite.py

[Tiny] Fix grammar in shared experts fusion log messages (#18043 )

2026-01-31 13:14:25 -08:00

glm4_moe_nextn.py

Use attn tp group in embedding for more models (#17570 )

2026-01-24 13:37:44 +08:00

glm4_moe.py

[NVIDIA] Add flashinfer all-to-all MOE dispatcher (#14668 )

2026-01-24 22:59:55 +08:00

glm4.py

[GLM-OCR] Support GLM-OCR Model (#17582 )

2026-01-26 22:24:00 -08:00

glm4v_moe.py

[Glm46v] Bug fix for accuracy drop and unable to launch server (#14585 )

2025-12-07 23:45:02 -08:00

glm4v.py

[GLM-OCR] Support GLM-OCR Model (#17582 )

2026-01-26 22:24:00 -08:00

glm_ocr_nextn.py

[GLM-OCR] Support GLM-OCR Model (#17582 )

2026-01-26 22:24:00 -08:00

glm_ocr.py

[GLM-OCR] Support GLM-OCR Model (#17582 )

2026-01-26 22:24:00 -08:00

glmasr.py

fix: change class name of GLM-ASR (#15772 )

2025-12-26 00:15:16 +08:00

gpt2.py

…

gpt_bigcode.py

…

gpt_j.py

[Fix] Remove unused Type import in gpt_j.py (#17975 )

2026-01-29 21:11:11 -08:00

gpt_oss.py

Use attn tp group in embedding for more models (#17570 )

2026-01-24 13:37:44 +08:00

granite.py

Fix incorrect default get_hidden_dim logic (#8987 )

2025-08-09 00:25:38 -07:00

granitemoe.py

[6/N] MoE Refactor: Cleanup MoE-related configs (#8849 )

2025-08-14 21:14:53 -07:00

grok.py

[Feature] Enable return routed experts (#12162 )

2025-12-21 15:16:43 +08:00

hunyuan.py

[Feature] Enable return routed experts (#12162 )

2025-12-21 15:16:43 +08:00

idefics2.py

Convert cu_seqlens to CPU for npu_flash_attention_unpad operator (#15434 )

2026-01-04 08:16:42 +08:00

internlm2_reward.py

…

internlm2.py

…

interns1.py

[Bug] Fix Intern-S1 model accuracy and support /generate interface with input_ids (#12367 )

2025-11-02 20:22:33 -08:00

internvl.py

[VLM] Support ViT CUDA Graph for InternVL (#16732 )

2026-01-14 17:29:23 +08:00

iquest_loopcoder.py

[Model] Support IQuest-Coder-40B-Loop (#16348 )

2026-01-12 23:44:45 +08:00

jet_nemotron.py

model: support JetVLM (#13289 )

2025-11-18 12:02:03 +08:00

jet_vlm.py

model: support JetVLM (#13289 )

2025-11-18 12:02:03 +08:00

kimi_k25.py

fix: avoid double reduce in VLM dp attention (#17991 )

2026-02-02 09:44:32 +08:00

kimi_linear.py

fuse qkvbfg linear into one gemm and f_b g_b into batched gemm. (#17801 )

2026-02-04 11:41:26 +08:00

kimi_vl_moonvit.py

move more files under srt/utils (#11285 )

2025-10-09 16:46:15 -07:00

kimi_vl.py

vlm: refactor engine vlm params and support processor output as input (#14091 )

2025-12-20 18:31:24 +08:00

lfm2.py

Add Liquid Foundation Model (LFM2) (#16890 )

2026-01-22 11:11:20 +08:00

lightonocr.py

Support LightOnOCR-2-1B (#17806 )

2026-01-29 23:03:41 +08:00

llada2.py

Use attn tp group in embedding for more models (#17570 )

2026-01-24 13:37:44 +08:00

llama4.py

add weightless qk norm to RMSNorm interface for Llama 4 (#12813 )

2026-01-29 19:09:55 -08:00

llama_classification.py

…

llama_eagle3.py

Eagle3 DP attention for Qwen3 MoE (#12002 )

2025-10-29 20:25:17 +08:00

llama_eagle.py

…

llama_embedding.py

…

llama_reward.py

…

llama.py

[Bugfix]Repeated add modelslim quant_config and bugfix with "enable-piecewise-cuda-graph" on NPU (#17511 )

2026-01-26 09:51:07 +08:00

llava.py

fix: Support PP for Mistral Small 3.1 (#14254 )

2025-12-02 13:04:14 +08:00

llavavid.py

refactor: unify names of the feature field of MultimodalDataItem (#8075 )

2025-07-16 17:52:38 -07:00

longcat_flash_nextn.py

Use attn tp group in embedding for more models (#17570 )

2026-01-24 13:37:44 +08:00

longcat_flash.py

Use attn tp group in embedding for more models (#17570 )

2026-01-24 13:37:44 +08:00

midashenglm.py

Add MiDasheng Model Support (#15219 )

2025-12-19 21:41:59 +08:00

mimo_mtp.py

[Lint] Add python/sglang to ruff F401 checks and remove unused imports in files (#11685 )

2025-10-17 16:49:46 -07:00

mimo_v2_flash_nextn.py

model: support Step-3.5-Flash (#18084 )

2026-02-03 00:40:07 +08:00

mimo_v2_flash.py

[MiMoV2Flash] [feat]: support two batch overlap (#17634 )

2026-02-02 14:03:57 -08:00

mimo.py

[Lint] Add python/sglang to ruff F401 checks and remove unused imports in files (#11685 )

2025-10-17 16:49:46 -07:00

mindspore.py

[NPU]mindspore model support moe (#15363 )

2026-02-02 17:52:49 +08:00

minicpm3.py

[NPU]support model MiniCPM3-4B for npu (#16866 )

2026-01-24 08:25:12 +08:00

minicpm.py

fix: remove redundant rotary embedding cache recomputation in MiniCPM (#8022 )

2025-07-15 00:12:45 -07:00

minicpmo.py

vlm: refactor engine vlm params and support processor output as input (#14091 )

2025-12-20 18:31:24 +08:00

minicpmv.py

[model] Support MiniCPM-V 4.5 (#9610 )

2026-01-31 23:37:36 -08:00

minimax_m2.py

Optimizing all_reduce in RMSNormTP in minimax_m2 (#16483 )

2026-01-31 23:39:37 -08:00

ministral3.py

ministral3 (#14251 )

2025-12-04 14:31:26 -08:00

mistral_large_3_eagle.py

Mistral Large 3 NVFP4 support (#14485 )

2025-12-13 13:34:42 +08:00

mistral_large_3.py

Mistral Large 3 NVFP4 support (#14485 )

2025-12-13 13:34:42 +08:00

mistral.py

refactor: unify names of the feature field of MultimodalDataItem (#8075 )

2025-07-16 17:52:38 -07:00

mixtral_quant.py

[feature]Ascend quantization support (#7791 )

2025-07-10 09:17:37 -07:00

mixtral.py

[Piecewise CUDA Graph] Fix recompile issue for Mixtral and Grok2 (#13667 )

2025-11-20 14:20:11 -08:00

mllama4.py

refactor: cleanup vision attention related codes (#13228 )

2025-11-16 22:44:40 +08:00

mllama.py

[Performance] Optimize Mllama LayerNorm -> Upd (#9725 )

2026-01-31 16:02:57 -08:00

nano_nemotron_vl.py

EVS Framework: Support NemotronH_Nano_VL_V2 (#14051 )

2026-01-05 16:18:07 +08:00

nemotron_h_mtp.py

Feat/support nemotron h mtp (#17013 )

2026-01-14 16:30:35 +08:00

nemotron_h.py

[NemotronH] Use ReplicatedLinear for fc1_latent_proj (#16569 )

2026-01-14 17:08:27 +08:00

nemotron_nas.py

from python.sglang.srt -> from sglang.srt (#9268 )

2025-08-17 02:45:45 -07:00

nvila_lite.py

model: support NVILA and NVILA Lite (#10399 )

2025-10-26 09:58:09 -07:00

nvila.py

model: support NVILA and NVILA Lite (#10399 )

2025-10-26 09:58:09 -07:00

olmo2.py

[perf] Add two stream norm for Olmo3 speedup 5% (#13681 )

2026-01-08 17:01:35 +00:00

olmo.py

support piecewise cuda graph for Olmo models (#14476 )

2025-12-06 17:57:44 -08:00

olmoe.py

[6/N] MoE Refactor: Cleanup MoE-related configs (#8849 )

2025-08-14 21:14:53 -07:00

opt.py

[lint] improve ruff check (#11922 )

2025-10-22 11:32:50 +08:00

orion.py

Support orion (#10665 )

2025-11-15 03:08:32 +08:00

paddleocr_vl.py

Convert cu_seqlens to CPU for npu_flash_attention_unpad operator (#15434 )

2026-01-04 08:16:42 +08:00

persimmon.py

Feat: Support for Persimmon Model (#7983 )

2025-07-19 23:07:47 -07:00

phi3_small.py

…

phi4mm_audio.py

Feat: Support audio in Phi4-mm model (#8048 )

2025-07-18 21:03:53 -07:00

phi4mm_utils.py

Feat: Support audio in Phi4-mm model (#8048 )

2025-07-18 21:03:53 -07:00

phi4mm.py

[Lint] Add python/sglang to ruff F401 checks and remove unused imports in files (#11685 )

2025-10-17 16:49:46 -07:00

phi.py

[Lint] Add python/sglang to ruff F401 checks and remove unused imports in files (#11685 )

2025-10-17 16:49:46 -07:00

phimoe.py

[Lint] Add python/sglang to ruff F401 checks and remove unused imports in files (#11685 )

2025-10-17 16:49:46 -07:00

pixtral.py

ministral3 (#14251 )

2025-12-04 14:31:26 -08:00

points_v15_chat.py

[model] Support POINTSV15Chat (#9651 )

2025-10-22 16:58:17 +08:00

qwen2_5_vl.py

[VLM] Support ViT CUDA Graph for InternVL (#16732 )

2026-01-14 17:29:23 +08:00

qwen2_audio.py

[Lint] Add python/sglang to ruff F401 checks and remove unused imports in files (#11685 )

2025-10-17 16:49:46 -07:00

qwen2_classification.py

…

qwen2_eagle.py

…

qwen2_moe.py

Use attn tp group in embedding for more models (#17570 )

2026-01-24 13:37:44 +08:00

qwen2_rm.py

…

qwen2_vl.py

Convert cu_seqlens to CPU for npu_flash_attention_unpad operator (#15434 )

2026-01-04 08:16:42 +08:00

qwen2.py

[PP] fix wrong weight logic for tie_word_embeddings model (#15890 )

2026-01-27 17:41:17 +08:00

qwen3_classification.py

Add support for Qwen3-seq-cls (#9357 )

2025-08-20 16:51:56 -07:00

qwen3_moe.py

[NPU] NZ for non-quantized MOE, Qwen3 MOE double memory consumption fix (#15904 )

2026-01-29 00:55:08 +08:00

qwen3_next_mtp.py

Depreate global_server_args_dict (#11528 )

2025-10-13 19:34:43 +08:00

qwen3_next.py

support qwen3-next eagle3 (#14607 )

2026-01-31 23:45:23 -08:00

qwen3_omni_moe.py

Convert cu_seqlens to CPU for npu_flash_attention_unpad operator (#15434 )

2026-01-04 08:16:42 +08:00

qwen3_vl_moe.py

[VLM] feat: true on policy for vlm + fsdp (#14636 )

2026-01-01 16:54:39 -08:00

qwen3_vl.py

fix: correct weight loading prefix mapping for Qwen3-VL (#18024 )

2026-02-01 21:50:21 -08:00

qwen3.py

Update weight rename check for Qwen3 Embeddings (#17535 )

2026-02-03 13:55:11 -08:00

qwen.py

[Lint] Add python/sglang to ruff F401 checks and remove unused imports in files (#11685 )

2025-10-17 16:49:46 -07:00

radio.py

Support nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-BF16 (and nvidia/C-RADIOv2-H) (#12277 )

2025-11-26 16:28:52 -07:00

registry.py

tiny unify environ usage (#15335 )

2025-12-17 23:31:43 +08:00

roberta.py

Implement BGE-M3 Sparse Embeddings in SGLang (#10869 )

2025-10-22 13:46:16 -07:00

sarashina2_vision.py

[Lint] Add python/sglang to ruff F401 checks and remove unused imports in files (#11685 )

2025-10-17 16:49:46 -07:00

siglip.py

refactor: cleanup vision attention related codes (#13228 )

2025-11-16 22:44:40 +08:00

solar.py

Remove wrong imports from sglang.python (#10493 )

2025-09-15 22:12:21 -07:00

stablelm.py

[NPU] solve accuracy problem for stablelm-2-1-6b for npu (#17470 )

2026-01-24 08:27:38 +08:00

starcoder2.py

model: support starcoder2 (#10609 )

2025-10-04 00:11:19 +08:00

step3_vl_10b.py

model: step3-vl-10b (#17513 )

2026-01-22 23:15:08 +08:00

step3_vl.py

Use attn tp group in embedding for more models (#17570 )

2026-01-24 13:37:44 +08:00

step3p5_mtp.py

model: support Step-3.5-Flash (#18084 )

2026-02-03 00:40:07 +08:00

step3p5.py

model: support Step-3.5-Flash (#18084 )

2026-02-03 00:40:07 +08:00

teleflm.py

model: support teleflm (#10573 )

2025-11-15 03:14:49 +08:00

torch_native_llama.py

Fix multimodal registry and code sync scripts (#10759 )

2025-09-22 15:36:01 -07:00

transformers.py

Small bug fix in transformers model implementation (#9809 )

2025-08-29 18:51:44 -07:00

utils.py

Use swa radix cache and memory pool for gpt-oss model (#17261 )

2026-01-18 13:47:07 +08:00

xverse_moe.py

[6/N] MoE Refactor: Cleanup MoE-related configs (#8849 )

2025-08-14 21:14:53 -07:00

xverse.py

…

yivl.py

…