baonudesifeizhai 54c8924384

[MoE Refactor][5/N] Isolate zero expert to LongCatFlash (#28891) ...

Signed-off-by: baonudesifeizhai <85092850+baonudesifeizhai@users.noreply.github.com>
Signed-off-by: Dongjie Zou <85092850+baonudesifeizhai@users.noreply.github.com>
Signed-off-by: baonudesifeizhai <baonudesifeizhai@gmail.com>
Signed-off-by: Robert Shaw <robertgshaw2@gmail.com>
Co-authored-by: Robert Shaw <robshaw@redhat.com>
Co-authored-by: Robert Shaw <robertgshaw2@gmail.com>

2025-12-20 18:22:04 +00:00

..

transformers

[ROCm][CI][Bugfix] Multi-Modal Model Support Fixes and Attention Backend Improvements (#30270)

2025-12-19 02:17:27 +00:00

__init__.py

…

adapters.py

[Model] Automatic conversion of TokenClassification model (#30666)

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

afmoe.py

[Bugfix][Model] Fix Afmoe rope_parameters issue (#30505)

2025-12-12 02:53:57 +00:00

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

…

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

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

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

audioflamingo3.py

Add AudioFlamingo3 model support (#30539)

2025-12-14 02:14:55 -08:00

aya_vision.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

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

bagel.py

[New Model] BAGEL support (AR only) (#28439)

2025-12-15 14:58:23 +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

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

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

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

…

bert_with_rope.py

…

bert.py

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

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

blip2.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

2025-12-06 07:57:28 +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

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

2025-12-19 02:04:19 +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

[Mamba] Removed disable cascade attn in MambaModelConfig (#30712)

2025-12-17 08:48:53 +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

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

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

deepseek_eagle.py

[Misc] Remove redundant attention var constants (#29650)

2025-11-28 04:35:19 -08:00

deepseek_mtp.py

[refactor] Add prefix support to embed_tokens in DeepSeek MTP (#30788)

2025-12-18 04:45:56 +00:00

deepseek_ocr.py

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

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

deepseek_v2.py

[Bug] Fix error 'Dynamo failed to run FX node with fake tensors for Deepseek V3.2 (#31046)

2025-12-19 23:31:31 +00:00

deepseek_vl2.py

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

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

dots1.py

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

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

dots_ocr.py

[CustomOp] Extract ApplyRotaryEmb as CustomOp and unify the dispatch logic (#29873)

2025-12-15 19:08:16 -08:00

ernie45_moe.py

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

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

ernie45_vl_moe.py

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

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

ernie45_vl.py

[CustomOp] Extract ApplyRotaryEmb as CustomOp and unify the dispatch logic (#29873)

2025-12-15 19:08:16 -08:00

ernie45.py

…

ernie_mtp.py

…

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

…

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

…

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

[CustomOp] Extract ApplyRotaryEmb as CustomOp and unify the dispatch logic (#29873)

2025-12-15 19:08:16 -08:00

glm4_moe_mtp.py

…

glm4_moe.py

GLM-4.7 Tool Parser and Doc Update (#30876)

2025-12-20 00:09:58 +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

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

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

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

gpt_oss.py

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

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

granite_speech.py

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

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

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

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

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

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

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

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

grok1.py

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

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

h2ovl.py

[Misc] Refactor tokenizer interface (#29693)

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

hunyuan_v1.py

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

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

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

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

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

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

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

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

interfaces_base.py

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

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

interfaces.py

[Misc] Improve error message for is_multimodal (#30483)

2025-12-11 06:39:51 -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

…

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

jais2.py

adds jais 2 support (#30188)

2025-12-18 15:46:58 +00:00

jais.py

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

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

jamba.py

[LoRA] Cleanup LoRA unused code (#29611)

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

jina_vl.py

Fix some Transformers nightly tests (#29802)

2025-12-02 07:05:27 +00:00

keye_vl1_5.py

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

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

keye.py

[CustomOp] Extract ApplyRotaryEmb as CustomOp and unify the dispatch logic (#29873)

2025-12-15 19:08:16 -08:00

kimi_linear.py

…

kimi_vl.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

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

lfm2_moe.py

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

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

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

[Chore] Deprecate merge_by_field_config arg (#30035)

2025-12-04 17:21:24 +00:00

llama4_eagle.py

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

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

llama4.py

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

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

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

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

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

llava_next_video.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

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

llava_next.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

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

llava_onevision.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

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

llava.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

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

longcat_flash_mtp.py

…

longcat_flash.py

[MoE Refactor][5/N] Isolate zero expert to LongCatFlash (#28891)

2025-12-20 18:22:04 +00:00

mamba2.py

…

mamba.py

…

medusa.py

…

midashenglm.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

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

mimo_mtp.py

…

mimo_v2_flash.py

[Model] Add MiMo-V2-Flash support (#30836)

2025-12-19 17:17:03 +00:00

mimo.py

…

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

[LoRA] Cleanup LoRA unused code (#29611)

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

minicpm.py

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

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

minicpmo.py

…

minicpmv.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

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

minimax_m2.py

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

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

minimax_text_01.py

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

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

minimax_vl_01.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

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

mistral3.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

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

mistral_large_3_eagle.py

[Deprecation] Remove fallbacks for embed_input_ids and embed_multimodal (#30458)

2025-12-11 06:58:23 +00:00

mistral_large_3.py

Add Mistral Large 3 and Ministral 3 (#29757)

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

mixtral.py

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

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

mllama4.py

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

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

mlp_speculator.py

…

modernbert.py

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

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

module_mapping.py

…

molmo.py

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

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

moonvit.py

[XPU]fix Kimi-VL-A3B-thinking on xpu (#29309)

2025-11-24 21:02:21 +00:00

mpt.py

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

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

nano_nemotron_vl.py

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

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

nemotron_h.py

Add latent MoE support (#30203)

2025-12-08 17:35:01 +00:00

nemotron_nas.py

Fix nemotron_nas intermediate_size computation (#30795)

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

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

…

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

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

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

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

…

openpangu.py

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

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

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

[CustomOp] Extract ApplyRotaryEmb as CustomOp and unify the dispatch logic (#29873)

2025-12-15 19:08:16 -08:00

paligemma.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

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

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

…

phi3v.py

[Misc] Improve error message for is_multimodal (#30483)

2025-12-11 06:39:51 -08:00

phi4mm_audio.py

…

phi4mm_utils.py

…

phi4mm.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

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

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

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

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

pixtral.py

[Chore] Adjust tokenizer import to avoid circular imports (#30601)

2025-12-13 04:42:39 -08:00

plamo2.py

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

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

plamo3.py

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

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

qwen2_5_omni_thinker.py

[Chore] Remove v0 dead code for Qwen2.5-omni (#30883)

2025-12-17 19:54:39 -08:00

qwen2_5_vl.py

[CustomOp] Extract ApplyRotaryEmb as CustomOp and unify the dispatch logic (#29873)

2025-12-15 19:08:16 -08:00

qwen2_audio.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

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

qwen2_moe.py

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

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

qwen2_rm.py

…

qwen2_vl.py

[CustomOp] Extract ApplyRotaryEmb as CustomOp and unify the dispatch logic (#29873)

2025-12-15 19:08:16 -08:00

qwen2.py

[New Model] BAGEL support (AR only) (#28439)

2025-12-15 14:58:23 +08:00

qwen3_moe.py

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

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

qwen3_next_mtp.py

…

qwen3_next.py

[Bugfix] Qwen3-next with --hf-overrides \{\"num_hidden_layers\":8\} (#30433)

2025-12-13 22:12:45 +08:00

qwen3_omni_moe_thinker.py

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

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

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][MM] Optimize encoder cache manager by operating with embeddings only (#30475)

2025-12-16 14:18:17 -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

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

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

qwen.py

[BugFix] Add embed_input_ids method to make QWenLMHeadModel a vllm model (#30674)

2025-12-15 10:38:29 +00:00

radio.py

…

registry.py

[Model] Add MiMo-V2-Flash support (#30836)

2025-12-19 17:17:03 +00:00

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

…

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

siglip2navit.py

[CustomOp] Extract ApplyRotaryEmb as CustomOp and unify the dispatch logic (#29873)

2025-12-15 19:08:16 -08:00

siglip.py

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

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

skyworkr1v.py

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

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

smolvlm.py

…

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

[Chore] Deprecate SupportsMultiModal.merge_by_field_config (#30170)

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

telechat2.py

…

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.py

[Chore] Adjust tokenizer import to avoid circular imports (#30601)

2025-12-13 04:42:39 -08:00

whisper.py

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

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

zamba2.py

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

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