Ce Zhaoand赵策 1123a87892 [Model] Enable LoRA support for Pixtral (#31724) ...

Signed-off-by: <>
Signed-off-by: 赵策 <alcor@zhaocedeMacBook-Air.local>
Signed-off-by: 赵策 <alcor@mac.mynetworksettings.com>
Co-authored-by: 赵策 <alcor@mac.mynetworksettings.com>

2026-01-08 05:00:57 -08:00

..

transformers

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

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

__init__.py

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

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

adapters.py

[Models] Allow converting Qwen3-VL into Reranker model (#31890)

2026-01-08 08:10:15 +00:00

afmoe.py

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

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

aimv2.py

[MM Encoder]: Migrate legacy ViT MultiHeadAttention to new MMEncoderAttention interface (#30684)

2025-12-19 02:04:19 +08:00

apertus.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

arcee.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

arctic.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

aria.py

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

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

audioflamingo3.py

[Bugfix] Enable dynamic_dims for different embeds shape (#31223)

2025-12-24 10:15:47 +08:00

aya_vision.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

2025-12-06 07:57:28 +00:00

bagel.py

[Bugfix] Fix BAGEL online serving for text and image understanding (#31546)

2025-12-31 14:46:10 -08:00

baichuan.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

bailing_moe.py

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

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

bamba.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

bee.py

[Model][VLM] Support Bee-8B Model (#27012)

2025-10-20 02:31:26 +00:00

bert_with_rope.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

bert.py

[Chore] Further cleanup pooler (#31951)

2026-01-08 02:16:21 -08:00

blip2.py

[Model] Enable LoRA support for BLIP2 (#31620)

2026-01-05 08:02:24 +00:00

blip.py

[MM Encoder]: Migrate legacy ViT MultiHeadAttention to new MMEncoderAttention interface (#30684)

2025-12-19 02:04:19 +08:00

bloom.py

[Attention] Remove imports from vllm/attention/__init__.py (#29342)

2025-11-26 10:53:15 -07:00

chameleon.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

chatglm.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

clip.py

[Model] Standardize common vision encoders (#31947)

2026-01-08 02:33:16 -08:00

cohere2_vision.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

2025-12-06 07:57:28 +00:00

commandr.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

config.py

[Models] Allow converting Qwen3-VL into Reranker model (#31890)

2026-01-08 08:10:15 +00:00

dbrx.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

deepencoder.py

[Model] Standardize common vision encoders (#31947)

2026-01-08 02:33:16 -08:00

deepseek_eagle.py

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

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

deepseek_mtp.py

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

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

deepseek_ocr.py

[Model] Standardize common vision encoders (#31947)

2026-01-08 02:33:16 -08:00

deepseek_v2.py

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

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

deepseek_vl2.py

Revert "[Renderer] Separate out RendererConfig from ModelConfig (#30145)" (#30199)

2025-12-07 00:00:22 -08:00

dots1.py

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

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

dots_ocr.py

[MM Encoder]: Make MMEncoderAttention's scale takes effect properly (#31950)

2026-01-08 02:33:48 -08:00

ernie45_moe.py

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

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

ernie45_vl_moe.py

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

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

ernie45_vl.py

[MM Encoder]: Make MMEncoderAttention's scale takes effect properly (#31950)

2026-01-08 02:33:48 -08:00

ernie45.py

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

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

ernie_mtp.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

exaone4.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

exaone.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

fairseq2_llama.py

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

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

falcon_h1.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

falcon.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

flex_olmo.py

Update Optional[x] -> x | None and Union[x, y] to x | y (#26633)

2025-10-12 09:51:31 -07:00

fuyu.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

2025-12-06 07:57:28 +00:00

gemma2.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

gemma3_mm.py

[Bugfix] Schedule failure due to wrong get_image_size_with_most_features (#29692)

2025-12-12 02:27:20 -08:00

gemma3.py

[v1] Add PrefixLM support to TritonAttention backend (#30386)

2025-12-17 16:05:24 -08:00

gemma3n_mm.py

Revert "[Renderer] Separate out RendererConfig from ModelConfig (#30145)" (#30199)

2025-12-07 00:00:22 -08:00

gemma3n.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

gemma.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

glm4_1v.py

[MM Encoder]: Make MMEncoderAttention's scale takes effect properly (#31950)

2026-01-08 02:33:48 -08:00

glm4_moe_mtp.py

[Bugfix][MTP] Fix GLM4 MoE fp8 loading with MTP on (#31757)

2026-01-07 09:18:52 +00:00

glm4_moe.py

[Model] Cleanup: Remove redundant manual definition of make_empty_intermediate_tensors in GLM-4-MoE (#31869)

2026-01-07 08:18:28 +00:00

glm4.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

glm4v.py

[MM Encoder]: Migrate legacy ViT MultiHeadAttention to new MMEncoderAttention interface (#30684)

2025-12-19 02:04:19 +08:00

glm.py

Access partial_rotary_factor from rope_parameters (#29966)

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

glmasr_utils.py

[Refactor] GLM-ASR Modeling (#31779)

2026-01-07 13:08:29 +00:00

glmasr.py

[MM Encoder]: Make MMEncoderAttention's scale takes effect properly (#31950)

2026-01-08 02:33:48 -08:00

gpt2.py

[Attention] Remove imports from vllm/attention/__init__.py (#29342)

2025-11-26 10:53:15 -07:00

gpt_bigcode.py

[Attention] Remove imports from vllm/attention/__init__.py (#29342)

2025-11-26 10:53:15 -07:00

gpt_j.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

gpt_neox.py

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

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

gpt_oss.py

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

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

granite_speech.py

[Misc] Improve error messages for unsupported types and parameters (#30593)

2026-01-07 09:00:16 +00:00

granite.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

granitemoe.py

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

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

granitemoehybrid.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

granitemoeshared.py

[LoRA] Cleanup LoRA unused code (#29611)

2025-11-28 22:52:58 -08:00

gritlm.py

[Refactor] Clean up pooler modules (#31897)

2026-01-08 00:07:43 +08:00

grok1.py

[Model] Add Grok-2 (#31847)

2026-01-08 04:59:48 -08:00

h2ovl.py

[Misc] Refactor tokenizer interface (#29693)

2025-11-29 04:02:21 -08:00

hunyuan_v1.py

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

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

hunyuan_vision.py

[MM Encoder]: Migrate legacy ViT MultiHeadAttention to new MMEncoderAttention interface (#30684)

2025-12-19 02:04:19 +08:00

hyperclovax_vision.py

[Model] Standardize common vision encoders (#31947)

2026-01-08 02:33:16 -08:00

idefics2_vision_model.py

[MM Encoder]: Migrate legacy ViT MultiHeadAttention to new MMEncoderAttention interface (#30684)

2025-12-19 02:04:19 +08:00

idefics3.py

[Core] Initialize LoRA support for tower and connector in multi-modal models (#26674)

2025-12-26 04:48:20 -08:00

interfaces_base.py

[Deprecation] Remove missed fallback for embed_input_ids (#30469)

2025-12-11 10:06:36 +00:00

interfaces.py

[Core] Initialize LoRA support for tower and connector in multi-modal models (#26674)

2025-12-26 04:48:20 -08:00

intern_vit.py

[MM Encoder]: Migrate legacy ViT MultiHeadAttention to new MMEncoderAttention interface (#30684)

2025-12-19 02:04:19 +08:00

internlm2_ve.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

internlm2.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

interns1_vit.py

[MM Encoder]: Migrate legacy ViT MultiHeadAttention to new MMEncoderAttention interface (#30684)

2025-12-19 02:04:19 +08:00

interns1.py

Revert "[Renderer] Separate out RendererConfig from ModelConfig (#30145)" (#30199)

2025-12-07 00:00:22 -08:00

internvl.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

2025-12-06 07:57:28 +00:00

isaac.py

[MM Encoder]: Make MMEncoderAttention's scale takes effect properly (#31950)

2026-01-08 02:33:48 -08:00

jais2.py

[Bugfix] Fix Jais2ForCausalLM (#31198)

2025-12-23 07:44:01 +00:00

jais.py

[Attention] Remove imports from vllm/attention/__init__.py (#29342)

2025-11-26 10:53:15 -07:00

jamba.py

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

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

jina_vl.py

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

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

keye_vl1_5.py

[Misc] Various cleanups for MM input processing (#29970)

2025-12-04 06:22:20 +00:00

keye.py

[Model] Standardize common vision encoders (#31947)

2026-01-08 02:33:16 -08:00

kimi_linear.py

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

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

kimi_vl.py

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

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

lfm2_moe.py

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

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

lfm2_vl.py

[Model] Add LFM2-VL model support (#31758)

2026-01-08 05:00:27 -08:00

lfm2.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

lightonocr.py

[Model] Standardize common vision encoders (#31947)

2026-01-08 02:33:16 -08:00

llama4_eagle.py

[Bugfix][llama4_eagle] Fix missing 'lm_head' attribute (#29926)

2025-12-05 19:57:26 +00:00

llama4.py

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

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

llama_eagle3.py

[Spec Decode] Add support for EAGLE3 heads that do not use_aux_hidden_states (#27688)

2025-11-24 20:13:12 -05:00

llama_eagle.py

[Bugfix] Fix spec decode memory regression after #28549 (#28819)

2025-11-20 19:05:50 +08:00

llama.py

[Docs] Improve malformed exception caused by backslash line continuations (#31694)

2026-01-05 17:51:54 -08:00

llava_next_video.py

[Model] Standardize common vision encoders (#31947)

2026-01-08 02:33:16 -08:00

llava_next.py

[Model] Standardize common vision encoders (#31947)

2026-01-08 02:33:16 -08:00

llava_onevision.py

[Model] Standardize common vision encoders (#31947)

2026-01-08 02:33:16 -08:00

llava.py

[Model] Standardize common vision encoders (#31947)

2026-01-08 02:33:16 -08:00

longcat_flash_mtp.py

Update Optional[x] -> x | None and Union[x, y] to x | y (#26633)

2025-10-12 09:51:31 -07:00

longcat_flash.py

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

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

mamba2.py

[Hybrid][torch.compile] Refactor mamba2 forward to avoid obscuring linear projections under custom op (#28587)

2025-11-18 16:49:36 -08:00

mamba.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

medusa.py

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

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

midashenglm.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

2025-12-06 07:57:28 +00:00

mimo_mtp.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

mimo_v2_flash.py

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

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

mimo.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

minicpm3.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

minicpm_eagle.py

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

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

minicpm.py

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

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

minicpmo.py

[Bugfix] Standardize merging multimodal embeddings (#26771)

2025-10-14 09:36:21 +00:00

minicpmv.py

[Bugfix] Enable dynamic_dims for different embeds shape (#31223)

2025-12-24 10:15:47 +08:00

minimax_m2.py

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

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

minimax_text_01.py

[Misc] Improve error messages for unsupported types and parameters (#30593)

2026-01-07 09:00:16 +00:00

minimax_vl_01.py

[Model] Standardize common vision encoders (#31947)

2026-01-08 02:33:16 -08:00

mistral3.py

[Model] Standardize common vision encoders (#31947)

2026-01-08 02:33:16 -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

mixtral.py

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

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

mllama4.py

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

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

mlp_speculator.py

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

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

modernbert.py

[Chore] Further cleanup pooler (#31951)

2026-01-08 02:16:21 -08:00

module_mapping.py

Update Optional[x] -> x | None and Union[x, y] to x | y (#26633)

2025-10-12 09:51:31 -07:00

molmo.py

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

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

moonvit.py

[MM Encoder]: Make MMEncoderAttention's scale takes effect properly (#31950)

2026-01-08 02:33:48 -08:00

mpt.py

[Attention] Remove imports from vllm/attention/__init__.py (#29342)

2025-11-26 10:53:15 -07:00

nano_nemotron_vl.py

[Model] Nemotron Parse 1.1 Support (#30864)

2026-01-05 13:00:14 -08:00

nemotron_h.py

Enable quantized attention in NemotronH models (#31898)

2026-01-07 17:37:19 +00:00

nemotron_nas.py

Fix nemotron_nas intermediate_size computation (#30795)

2025-12-17 01:06:28 +00:00

nemotron_parse.py

[Model] Nemotron Parse 1.1 Support (#30864)

2026-01-05 13:00:14 -08:00

nemotron_vl.py

Revert "[Renderer] Separate out RendererConfig from ModelConfig (#30145)" (#30199)

2025-12-07 00:00:22 -08:00

nemotron.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

nvlm_d.py

Update Optional[x] -> x | None and Union[x, y] to x | y (#26633)

2025-10-12 09:51:31 -07:00

olmo2.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

olmo.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

olmoe.py

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

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

opencua.py

[CustomOp][MM] Extract MMEncoderAttention as CustomOp and replace the backend of QwenVisionAttention with it. (#30125)

2025-12-15 11:13:32 +08:00

openpangu_mtp.py

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

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

openpangu.py

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

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

opt.py

[Attention] Remove imports from vllm/attention/__init__.py (#29342)

2025-11-26 10:53:15 -07:00

orion.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

ouro.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

ovis2_5.py

[CustomOp][MM] Extract MMEncoderAttention as CustomOp and replace the backend of QwenVisionAttention with it. (#30125)

2025-12-15 11:13:32 +08:00

ovis.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

2025-12-06 07:57:28 +00:00

paddleocr_vl.py

[MM Encoder]: Make MMEncoderAttention's scale takes effect properly (#31950)

2026-01-08 02:33:48 -08:00

paligemma.py

[Model] Enable LoRA support for PaliGemma (#31656)

2026-01-07 10:09:32 +08:00

persimmon.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

phi3.py

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

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

phi3v.py

[Model] Standardize common vision encoders (#31947)

2026-01-08 02:33:16 -08:00

phi4mm_audio.py

Update Optional[x] -> x | None and Union[x, y] to x | y (#26633)

2025-10-12 09:51:31 -07:00

phi4mm_utils.py

Update Optional[x] -> x | None and Union[x, y] to x | y (#26633)

2025-10-12 09:51:31 -07:00

phi4mm.py

[Docs] Improve malformed exception caused by backslash line continuations (#31694)

2026-01-05 17:51:54 -08:00

phi.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

phimoe.py

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

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

pixtral.py

[Model] Enable LoRA support for Pixtral (#31724)

2026-01-08 05:00:57 -08:00

plamo2.py

Support LoRA and GPTQModel for PLaMo 2/3 (#31322)

2025-12-26 11:41:33 +08:00

plamo3.py

Support LoRA and GPTQModel for PLaMo 2/3 (#31322)

2025-12-26 11:41:33 +08:00

qwen2_5_omni_thinker.py

[Doc] Show that use_audio_in_video is supported in docs (#30837)

2026-01-05 23:27:19 -08:00

qwen2_5_vl.py

[MM Encoder]: Make MMEncoderAttention's scale takes effect properly (#31950)

2026-01-08 02:33:48 -08:00

qwen2_audio.py

[Bugfix] Enable dynamic_dims for different embeds shape (#31223)

2025-12-24 10:15:47 +08:00

qwen2_moe.py

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

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

qwen2_rm.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

qwen2_vl.py

[MM Encoder]: Make MMEncoderAttention's scale takes effect properly (#31950)

2026-01-08 02:33:48 -08:00

qwen2.py

[Cleanup] Remove redundant decoder_layer_type assignment in Qwen2 (#31760)

2026-01-05 18:09:18 -08:00

qwen3_moe.py

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

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

qwen3_next_mtp.py

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

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

qwen3_next.py

[Misc] Support qwen3-next lora (#31719)

2026-01-08 09:27:50 +00:00

qwen3_omni_moe_thinker.py

[Bugfix]: avoid overriding audio/text kwargs (Qwen3-Omni) (#31790)

2026-01-06 12:59:17 +00:00

qwen3_vl_moe.py

[Bugfix] Fix multimodal configuration for Qwen3VL MOE model (#30670)

2025-12-15 14:06:01 +00:00

qwen3_vl.py

[Core] Initialize LoRA support for tower and connector in multi-modal models (#26674)

2025-12-26 04:48:20 -08:00

qwen3.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

qwen_vl.py

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

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

qwen.py

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

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

radio.py

[Model] Nemotron Parse 1.1 Support (#30864)

2026-01-05 13:00:14 -08:00

registry.py

[Model] Add LFM2-VL model support (#31758)

2026-01-08 05:00:27 -08:00

roberta.py

Don'e assume position_embedding_type will be present for BERT and RoBERTa models (#30770)

2025-12-16 13:40:26 +00:00

rvl.py

Update Optional[x] -> x | None and Union[x, y] to x | y (#26633)

2025-10-12 09:51:31 -07:00

seed_oss.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

siglip2.py

[Model] Add LFM2-VL model support (#31758)

2026-01-08 05:00:27 -08:00

siglip2navit.py

[Model] Standardize common vision encoders (#31947)

2026-01-08 02:33:16 -08:00

siglip.py

[Model] Standardize common vision encoders (#31947)

2026-01-08 02:33:16 -08:00

skyworkr1v.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

2025-12-06 07:57:28 +00:00

smolvlm.py

Update Optional[x] -> x | None and Union[x, y] to x | y (#26633)

2025-10-12 09:51:31 -07:00

solar.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

stablelm.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

starcoder2.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

step3_text.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

step3_vl.py

[MM Encoder]: Migrate legacy ViT MultiHeadAttention to new MMEncoderAttention interface (#30684)

2025-12-19 02:04:19 +08:00

swin.py

Remove head_mask from Ultravox and Swin (#30764)

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

tarsier.py

[Model] Standardize common vision encoders (#31947)

2026-01-08 02:33:16 -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

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

2025-12-06 07:57:28 +00:00

ultravox.py

Remove head_mask from Ultravox and Swin (#30764)

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

utils.py

[Deprecation] Advance deprecation status (#29617)

2025-11-27 19:04:58 +00:00

vision.py

[Bugfix] Fix broken ViT attention selection for Blackwell device (#30731)

2025-12-16 05:24:32 +00:00

voxtral_streaming.py

[Voxtral] Fix speech transcription api (#31388)

2026-01-08 18:34:19 +08:00

voxtral.py

[Voxtral] Fix speech transcription api (#31388)

2026-01-08 18:34:19 +08:00

whisper_utils.py

adapt voxtral (#31095)

2025-12-23 05:31:55 -08:00

whisper.py

[Voxtral] Fix speech transcription api (#31388)

2026-01-08 18:34:19 +08:00

zamba2.py

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

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