Max Hu 9c3fe9936b Flashinfer cuDNN backend for Qwen3 VL ViT attention (#34580) ...

Signed-off-by: Max Hu <maxhu@nvidia.com>
Signed-off-by: Max Hu <hyoung2991@gmail.com>
Co-authored-by: Max Hu <maxhu@nvidia.com>
Co-authored-by: Shang Wang <shangw@nvidia.com>

2026-02-27 20:20:23 +08:00

..

transformers

[Refactor] Decouple TimingContext from InputProcessingContext (#35083)

2026-02-23 14:15:50 +00:00

__init__.py

[Chore] Remove SupportsV0Only interface and update supported models docs (#26783)

2025-10-14 04:55:10 -07:00

adapters.py

[Bugfix] Fix Qwen3-VL-Reranker load. (#33298)

2026-01-29 08:42:53 +00:00

afmoe.py

[Bugfix] Define router_logits_dtype for remaining MoE models (#33737)

2026-02-04 13:24:14 +08:00

aimv2.py

[MM] Align the prefix of MMEncoderAttention with Attention (#33750)

2026-02-04 04:07:30 +00:00

apertus.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

arcee.py

[Chore] Update type annotation of input_ids in model forward (#33063)

2026-01-26 06:02:10 -08:00

arctic.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

aria.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

audioflamingo3.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

aya_vision.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

bagel.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

baichuan.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

bailing_moe_linear.py

[Model] Ring 2.5 (#35102)

2026-02-26 02:17:11 -08:00

bailing_moe.py

[Bugfix] Define router_logits_dtype for remaining MoE models (#33737)

2026-02-04 13:24:14 +08:00

bamba.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

bee.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

bert_with_rope.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

bert.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

blip2.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

blip.py

[MM] Align the prefix of MMEncoderAttention with Attention (#33750)

2026-02-04 04:07:30 +00:00

bloom.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

chameleon.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

chatglm.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

clip.py

[Refactor] Decouple TimingContext from InputProcessingContext (#35083)

2026-02-23 14:15:50 +00:00

cohere2_vision.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

colbert.py

Extend ColBERT support to non-standard BERT backbones (#34170)

2026-02-13 09:53:09 +00:00

colmodernvbert.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

colqwen3.py

[Frontend] Support multimodal inputs for late-interaction scoring (ColQwen3) + NewModel: nvidia/nemotron-colembed (#34574)

2026-02-20 20:01:40 -08:00

commandr.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

config.py

[Bug] Fix outdated links in source code (#35314)

2026-02-27 03:50:46 +00:00

dbrx.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

deepencoder2.py

[Models] Consolidate Deepseek-OCR2 processor (#33909)

2026-02-05 18:29:20 +00:00

deepencoder.py

[Model] Support DeepSeek-OCR-2 (#33165)

2026-02-02 06:24:10 +00:00

deepseek_eagle.py

[BugFix] LoRA: Support loading base_layer of experts (#31104)

2026-01-07 14:49:39 +08:00

deepseek_mtp.py

[ROCm][Bugfix] Fix ptpc scale load issue for fused shared expert path in deepseek mtp (#33018)

2026-01-26 23:19:04 +08:00

deepseek_ocr2.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

deepseek_ocr.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

deepseek_v2.py

[Performance] Cublas Bf16 Gate with Fp32 Output (#35121)

2026-02-26 16:51:28 -08:00

deepseek_vl2.py

[Refactor] Decouple TimingContext from InputProcessingContext (#35083)

2026-02-23 14:15:50 +00:00

dots1.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

dots_ocr.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

eagle2_5_vl.py

[Model] Apply #32631 for recent models (#33785)

2026-02-04 12:23:01 +00:00

ernie45_moe.py

Remove padding_index from models that don't use it for better Transformers v5 compatibility (#35189)

2026-02-24 08:04:46 -08:00

ernie45_vl_moe.py

Remove padding_index from models that don't use it for better Transformers v5 compatibility (#35189)

2026-02-24 08:04:46 -08:00

ernie45_vl.py

[Misc] Standardize handling of mm_processor_kwargs.size (#35284)

2026-02-26 13:05:46 +00:00

ernie45.py

[Model] Add PaddleOCR-VL Model Support (#27758)

2025-11-03 19:04:22 +08:00

ernie_mtp.py

[Chore] Update type annotation of input_ids in model forward (#33063)

2026-01-26 06:02:10 -08:00

exaone4.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

exaone_moe_mtp.py

Add K-EXAONE-236B-A23B (#31621)

2026-01-12 16:30:50 +00:00

exaone_moe.py

[Chore] Update type annotation of input_ids in model forward (#33063)

2026-01-26 06:02:10 -08:00

exaone.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

fairseq2_llama.py

Convert formatting to use ruff instead of yapf + isort (#26247)

2025-10-05 07:06:22 -07:00

falcon_h1.py

Fix quantized Falcon-H1 model loading issues (#32728)

2026-02-02 22:31:27 -08:00

falcon.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

flex_olmo.py

[Bugfix] Define router_logits_dtype for remaining MoE models (#33737)

2026-02-04 13:24:14 +08:00

funasr.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

funaudiochat.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

fuyu.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

gemma2.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

gemma3_mm.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

gemma3.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

gemma3n_audio_utils.py

[Bugfix] Fix integer overflow in Gemma3n audio processing (#31657)

2026-01-10 17:52:53 +08:00

gemma3n_mm.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

gemma3n.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

gemma.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

glm4_1v.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

glm4_moe_lite_mtp.py

[Chore] Update type annotation of input_ids in model forward (#33063)

2026-01-26 06:02:10 -08:00

glm4_moe_lite.py

Update get_expert_mapping to include self parameter (#33525)

2026-02-02 20:29:07 +08:00

glm4_moe_mtp.py

[Chore] Update type annotation of input_ids in model forward (#33063)

2026-01-26 06:02:10 -08:00

glm4_moe.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

glm4.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

glm4v.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

glm_ocr_mtp.py

[GLM-OCR] GLM-OCR with MTP Support (#33005)

2026-01-26 06:24:43 -08:00

glm_ocr.py

[MM] Align the prefix of MMEncoderAttention with Attention (#33750)

2026-02-04 04:07:30 +00:00

glm.py

Access partial_rotary_factor from rope_parameters (#29966)

2025-12-04 18:42:49 +00:00

glmasr_utils.py

[Refactor] Remove dead private func _fp8_perm and _extract_mask_for_item (#35068)

2026-02-23 05:05:20 -08:00

glmasr.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

gpt2.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

gpt_bigcode.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

gpt_j.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

gpt_neox.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

gpt_oss.py

fix(docs): fix typos in comments and docstrings (#34836)

2026-02-18 23:17:41 -08:00

granite_speech.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

granite.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

granitemoe.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

granitemoehybrid.py

fix: Apply embedding_multiplier to inputs_embeds (#34813)

2026-02-23 00:42:46 -08:00

granitemoeshared.py

Remove padding_index from models that don't use it for better Transformers v5 compatibility (#35189)

2026-02-24 08:04:46 -08:00

gritlm.py

[Model] Standardize pooling heads (#32148)

2026-01-12 17:01:49 +00:00

grok1.py

Remove padding_index from models that don't use it for better Transformers v5 compatibility (#35189)

2026-02-24 08:04:46 -08:00

h2ovl.py

[Refactor] Decouple TimingContext from InputProcessingContext (#35083)

2026-02-23 14:15:50 +00:00

hunyuan_v1.py

Remove padding_index from models that don't use it for better Transformers v5 compatibility (#35189)

2026-02-24 08:04:46 -08:00

hunyuan_vision.py

[Misc] Standardize handling of mm_processor_kwargs.size (#35284)

2026-02-26 13:05:46 +00:00

hyperclovax_vision.py

[Refactor] Remove dead or duplicate func utils or variables (#35318)

2026-02-26 10:57:56 -05:00

idefics2_vision_model.py

[MM] Align the prefix of MMEncoderAttention with Attention (#33750)

2026-02-04 04:07:30 +00:00

idefics3.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

interfaces_base.py

[Refactor] Separate sequence and token pooling types (#32026)

2026-01-10 04:53:24 +00:00

interfaces.py

[Frontend] Add automatic language detection for Whisper transcription (#34342)

2026-02-21 04:49:41 -08:00

intern_vit.py

[MM] Align the prefix of MMEncoderAttention with Attention (#33750)

2026-02-04 04:07:30 +00:00

internlm2_ve.py

[Chore] Update type annotation of input_ids in model forward (#33063)

2026-01-26 06:02:10 -08:00

internlm2.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

interns1_pro.py

[CI/Build] Enable tests for recent day-0 new models (#34585)

2026-02-15 18:17:04 -08:00

interns1_vit.py

[MM] Align the prefix of MMEncoderAttention with Attention (#33750)

2026-02-04 04:07:30 +00:00

interns1.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

internvl.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

iquest_loopcoder.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

isaac.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

jais2.py

Remove padding_index from models that don't use it for better Transformers v5 compatibility (#35189)

2026-02-24 08:04:46 -08:00

jais.py

Make JAIS compatible with Transformers v5 (#34264)

2026-02-11 12:30:37 +00:00

jamba.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

jina_vl.py

[Chore] Update type annotation of input_ids in model forward (#33063)

2026-01-26 06:02:10 -08:00

kanana_v.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

keye_vl1_5.py

[Refactor] Define MM data parser in processing info instead of processor itself (#33260)

2026-01-29 13:55:17 +08:00

keye.py

[Misc] Standardize handling of mm_processor_kwargs.size (#35284)

2026-02-26 13:05:46 +00:00

kimi_k25_vit.py

[Bugfix] Add quant_config in ViT of Kimi-K2.5 (#34501)

2026-02-13 16:05:34 +00:00

kimi_k25.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

kimi_linear.py

Remove padding_index from models that don't use it for better Transformers v5 compatibility (#35189)

2026-02-24 08:04:46 -08:00

kimi_vl.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

lfm2_moe.py

[Models] LFM2: Support LoRA (#34921)

2026-02-19 22:07:23 -08:00

lfm2_siglip2.py

[Models]: lfm2_siglip2 return intermediate encoder layers (#33370)

2026-02-01 06:17:49 +00:00

lfm2_vl.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

lfm2.py

[Models] LFM2: Support LoRA (#34921)

2026-02-19 22:07:23 -08:00

lightonocr.py

[Models]: Make Multimodal config implicit in ViT implementation (#31972)

2026-01-24 20:34:26 +08:00

llama4_eagle.py

[Spec Decode] Reduce TP communication for speculative decoding draft token generation (#34049)

2026-02-22 14:59:16 -08:00

llama4.py

[Llama4,CI] Bring back Llama-4 bug fixes, and also fix Maverick tests (#35033)

2026-02-23 09:04:34 -05:00

llama_eagle3.py

[Spec Decode] Unified Parallel Drafting (#32887)

2026-02-05 12:37:18 -05:00

llama_eagle.py

Add llmcompressor fp8 kv-cache quant (per-tensor and per-attn_head) (#30141)

2026-01-22 13:29:57 -07:00

llama.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

llava_next_video.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

llava_next.py

[Bugfix] Add Quant Config to Llava Next Projector (#34847)

2026-02-18 23:18:23 -08:00

llava_onevision.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

llava.py

[Refactor] Decouple TimingContext from InputProcessingContext (#35083)

2026-02-23 14:15:50 +00:00

longcat_flash_mtp.py

[Chore] Update type annotation of input_ids in model forward (#33063)

2026-01-26 06:02:10 -08:00

longcat_flash.py

Remove padding_index from models that don't use it for better Transformers v5 compatibility (#35189)

2026-02-24 08:04:46 -08:00

mamba2.py

[Model][Spec Decode] Nemotron-H MTP and Mamba Speculative Decoding Support (#33726)

2026-02-24 09:49:56 -08:00

mamba.py

[Chore] Update type annotation of input_ids in model forward (#33063)

2026-01-26 06:02:10 -08:00

medusa.py

[LoRA][1/N]Remove LoRA extra vocab (#28382)

2025-11-11 11:06:21 -08:00

midashenglm.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

mimo_mtp.py

[Chore] Update type annotation of input_ids in model forward (#33063)

2026-01-26 06:02:10 -08:00

mimo_v2_flash.py

[Bugfix] Define router_logits_dtype for remaining MoE models (#33737)

2026-02-04 13:24:14 +08:00

mimo.py

[Chore] Update type annotation of input_ids in model forward (#33063)

2026-01-26 06:02:10 -08:00

minicpm3.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

minicpm_eagle.py

[Misc][Model][Refactor] Pass the prefix into Linear layers (#31669)

2026-01-05 20:03:18 +00:00

minicpm.py

[Model Bash][DeepSeekR1] Remove Shared Expert Clone (#34344)

2026-02-19 07:56:14 -08:00

minicpmo.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

minicpmv.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

minimax_m2.py

[Bugfix] Disable TRTLLM FP8 MoE if router_logits_dtype==float32 and routing_method!=DeepSeekV3 (#33613)

2026-02-03 13:26:51 -08:00

minimax_text_01.py

Remove padding_index from models that don't use it for better Transformers v5 compatibility (#35189)

2026-02-24 08:04:46 -08:00

minimax_vl_01.py

[Bugfix] Fix models and tests for transformers v5 (#33977)

2026-02-06 21:47:41 +08:00

mistral3.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

mistral_large_3_eagle.py

[Misc][Model][Refactor] Pass the prefix into Linear layers (#31669)

2026-01-05 20:03:18 +00:00

mistral_large_3.py

Add Mistral Large 3 and Ministral 3 (#29757)

2025-12-02 10:29:00 +00:00

mistral.py

Remove unused logic in models/mistral.py (#33095)

2026-01-26 09:01:52 -08:00

mixtral.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

mllama4.py

[Feature] Add LoRA tower/connector support for Llama 4 Vision (mllama4) (#35147)

2026-02-24 04:10:32 -08:00

mlp_speculator.py

[LoRA][1/N]Remove LoRA extra vocab (#28382)

2025-11-11 11:06:21 -08:00

modernbert.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

module_mapping.py

[Misc] Remove unused ModelKeys (#32608)

2026-01-19 17:34:59 +00:00

molmo2.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

molmo.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

moonvit.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

mpt.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

musicflamingo.py

[Model][Multimodal] Add explicit MusicFlamingo adapter (#32696)

2026-01-30 11:01:29 +08:00

nano_nemotron_vl.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

nemotron_h_mtp.py

[Model][Spec Decode] Nemotron-H MTP and Mamba Speculative Decoding Support (#33726)

2026-02-24 09:49:56 -08:00

nemotron_h.py

[Performance] Cublas Bf16 Gate with Fp32 Output (#35121)

2026-02-26 16:51:28 -08:00

nemotron_nas.py

Remove padding_index from models that don't use it for better Transformers v5 compatibility (#35189)

2026-02-24 08:04:46 -08:00

nemotron_parse.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

nemotron_vl.py

[Model] Add nvidia/llama-nemotron-embed-vl-1b-v2 multimodal embedding model (#35297)

2026-02-26 14:17:17 +00:00

nemotron.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

nvlm_d.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

olmo2.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

olmo.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

olmoe.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

opencua.py

[Refactor] Define MM data parser in processing info instead of processor itself (#33260)

2026-01-29 13:55:17 +08:00

openpangu_mtp.py

[Chore] Update type annotation of input_ids in model forward (#33063)

2026-01-26 06:02:10 -08:00

openpangu_vl.py

[Model] Apply #32631 for recent models (#33785)

2026-02-04 12:23:01 +00:00

openpangu.py

Remove padding_index from models that don't use it for better Transformers v5 compatibility (#35189)

2026-02-24 08:04:46 -08:00

opt.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

orion.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

ouro.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

ovis2_5.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

ovis.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

paddleocr_vl.py

[Misc] Standardize handling of mm_processor_kwargs.size (#35284)

2026-02-26 13:05:46 +00:00

paligemma.py

[Refactor] Decouple TimingContext from InputProcessingContext (#35083)

2026-02-23 14:15:50 +00:00

persimmon.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

phi3.py

Convert formatting to use ruff instead of yapf + isort (#26247)

2025-10-05 07:06:22 -07:00

phi3v.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

phi4mm_audio.py

[Refactor] Clean up unused variables & func (#32692)

2026-01-23 17:04:25 -05:00

phi4mm_utils.py

[Chore] Replace swish with silu (#32459)

2026-01-16 08:22:45 +00:00

phi4mm.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

phi.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

phimoe.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

pixtral.py

[Refactor] Decouple TimingContext from InputProcessingContext (#35083)

2026-02-23 14:15:50 +00:00

plamo2.py

[Model][Spec Decode] Nemotron-H MTP and Mamba Speculative Decoding Support (#33726)

2026-02-24 09:49:56 -08:00

plamo3.py

Remove padding_index from models that don't use it for better Transformers v5 compatibility (#35189)

2026-02-24 08:04:46 -08:00

qwen2_5_omni_thinker.py

[Bugfix] Fix Qwen2.5-Omni and Qwen3-Omni mixed-modality embed regression (#35368)

2026-02-26 11:58:23 +00:00

qwen2_5_vl.py

Flashinfer cuDNN backend for Qwen3 VL ViT attention (#34580)

2026-02-27 20:20:23 +08:00

qwen2_audio.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

qwen2_moe.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

qwen2_rm.py

[Chore] Update type annotation of input_ids in model forward (#33063)

2026-01-26 06:02:10 -08:00

qwen2_vl.py

[Misc] Standardize handling of mm_processor_kwargs.size (#35284)

2026-02-26 13:05:46 +00:00

qwen2.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

qwen3_5_mtp.py

[Misc] Port Qwen3.5 Configs (#34512)

2026-02-13 05:24:25 -08:00

qwen3_5.py

[Bugfix] Qwen3.5 kv-scale weight remapping (#34719)

2026-02-19 04:13:37 -08:00

qwen3_asr_realtime.py

[Realtime] Add Qwen3-ASR realtime streaming support (#34613)

2026-02-20 19:59:42 -08:00

qwen3_asr.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

qwen3_moe.py

Remove padding_index from models that don't use it for better Transformers v5 compatibility (#35189)

2026-02-24 08:04:46 -08:00

qwen3_next_mtp.py

[Chore] Update type annotation of input_ids in model forward (#33063)

2026-01-26 06:02:10 -08:00

qwen3_next.py

[Bugfix] Fix Qwen3NextForCausalLM packed_modules_mapping (#35413)

2026-02-26 19:46:30 -08:00

qwen3_omni_moe_thinker.py

[Bugfix] Fix Qwen2.5-Omni and Qwen3-Omni mixed-modality embed regression (#35368)

2026-02-26 11:58:23 +00:00

qwen3_vl_moe.py

Make Qwen3VL compatible with Transformers v5 (#34262)

2026-02-11 04:13:23 -08:00

qwen3_vl.py

Flashinfer cuDNN backend for Qwen3 VL ViT attention (#34580)

2026-02-27 20:20:23 +08:00

qwen3.py

[Bugfix][Model] Support LoRA on Qwen3 Output Embedding (#29816)

2026-02-06 20:25:31 +08:00

qwen_vl.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

qwen.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

radio.py

[5/N] Initialize MM components in context managers (Q-Z) (#32695)

2026-01-20 19:10:23 +00:00

registry.py

[Model] Add nvidia/llama-nemotron-embed-vl-1b-v2 multimodal embedding model (#35297)

2026-02-26 14:17:17 +00:00

roberta.py

Support bge-m3 sparse embeddings and colbert embeddings (#14526)

2026-01-22 23:52:57 +08:00

rvl.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

seed_oss.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

siglip2navit.py

[New Model] support new model ovis2.6 (#34426)

2026-02-13 00:12:45 -08:00

siglip.py

[Refactor] Decouple TimingContext from InputProcessingContext (#35083)

2026-02-23 14:15:50 +00:00

skyworkr1v.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

smolvlm.py

[Bugfix] Standardize getting number of image patches/tokens (#34358)

2026-02-12 20:47:01 -08:00

solar.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

stablelm.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

starcoder2.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

step1.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

step3_text.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00

step3_vl.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

step3p5_mtp.py

[Models] Step-3.5-Flash (#33523)

2026-02-02 10:21:18 +08:00

step3p5.py

[Model] Add NVFP4 quantization support for Step3.5-Flash (#34478)

2026-02-22 12:30:46 -07:00

step_vl.py

[MM] Align the prefix of MMEncoderAttention with Attention (#33750)

2026-02-04 04:07:30 +00:00

swin.py

Remove head_mask from Ultravox and Swin (#30764)

2025-12-16 08:02:41 -08:00

tarsier.py

[Chore] Update type annotation of input_ids in model forward (#33063)

2026-01-26 06:02:10 -08:00

telechat2.py

Convert formatting to use ruff instead of yapf + isort (#26247)

2025-10-05 07:06:22 -07:00

teleflm.py

[LoRA][2/2]Remove LoRA extra vocab (#28545)

2025-11-21 09:46:43 +08:00

terratorch.py

[Refactor] Decouple TimingContext from InputProcessingContext (#35083)

2026-02-23 14:15:50 +00:00

ultravox.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

utils.py

[offloader] v2: Hide weight onloading latency via prefetching (#29941)

2026-02-25 17:20:59 -08:00

vision.py

[MM Encoder] Add Triton ViT attention backend (#32183)

2026-02-15 06:32:47 -08:00

voxtral_realtime.py

Make voxtral compile friendly (#33959)

2026-02-24 09:33:35 +01:00

voxtral.py

[Refactor] Decouple TimingContext from InputProcessingContext (#35083)

2026-02-23 14:15:50 +00:00

voyage.py

Fix main pre-commit (#33975)

2026-02-06 00:08:05 -08:00

whisper_causal.py

[CI] Heavy refactoring of Voxtral multimodal audio model tests (#34294)

2026-02-13 20:04:29 -08:00

whisper_utils.py

[Voxtral] Add new streaming arch (#32861)

2026-01-23 12:41:52 +01:00

whisper.py

[Refactor] Simplify dummy data generation (#35025)

2026-02-22 20:55:27 -08:00

zamba2.py

[5/N][Attention] Finish eliminating vllm/attention folder (#32064)

2026-01-27 10:02:51 -05:00