🔴 [VLM] Add base model without head (#37033)

* i guessreverted all CdGen classes

* style

* llava onevision

* fix copies

* fix some tests

* some more tests

* dump

* skip these

* nevermind, i am dumb

* revert fix not needed

* fixup

* fixup

* another fixup

* more fixup to make ci finally happy

* fixup after rebasing

* fix qwen tests

* add internVL + typos here and there

* image token index -> id

* style

* fix init weights

* revert blip-2 not supported

* address comments

* fix copies

* revert blip2 test file as well

* as discussed internally, revert back CdGen models

* fix some tests

* fix more tests for compile

* CI red

* fix copies

* enumerate explicitly allowed models

* address comments

* fix tests

* fixup

* style again

* add tests for new model class

* another fixup ( x _ x )

* [fixup] unused attributes can be removed post-deprecation

docs/source/en/model_doc/aria.md

@@ -102,6 +102,10 @@ response = processor.decode(output_ids, skip_special_tokens=True)
 
 [[autodoc]] AriaTextModel
 
+## AriaModel
+
+[[autodoc]] AriaModel
+
 ## AriaTextForCausalLM
 
 [[autodoc]] AriaTextForCausalLM

docs/source/en/model_doc/aya_vision.md

@@ -237,6 +237,10 @@ for i, output in enumerate(batch_outputs):
 
 [[autodoc]] AyaVisionConfig
 
+## AyaVisionModel
+
+[[autodoc]] AyaVisionModel
+
 ## AyaVisionForConditionalGeneration
 
 [[autodoc]] AyaVisionForConditionalGeneration

docs/source/en/model_doc/emu3.md

@@ -174,6 +174,10 @@ for i, image in enumerate(images['pixel_values']):
 [[autodoc]] Emu3TextModel
     - forward
 
+## Emu3Model
+
+[[autodoc]] Emu3Model
+
 ## Emu3ForCausalLM
 
 [[autodoc]] Emu3ForCausalLM

docs/source/en/model_doc/fuyu.md

@@ -103,6 +103,10 @@ The `LlamaTokenizer` is used as it is a standard wrapper around sentencepiece.
 
 [[autodoc]] FuyuConfig
 
+## FuyuModel
+
+[[autodoc]] FuyuModel
+
 ## FuyuForCausalLM
 
 [[autodoc]] FuyuForCausalLM

docs/source/en/model_doc/gemma3.md

@@ -254,6 +254,10 @@ visualizer("<img>What is shown in this image?")
 [[autodoc]] Gemma3TextModel
     - forward
 
+## Gemma3Model
+
+[[autodoc]] Gemma3Model
+
 ## Gemma3ForCausalLM
 
 [[autodoc]] Gemma3ForCausalLM

docs/source/en/model_doc/got_ocr2.md

@@ -277,6 +277,10 @@ alt="drawing" width="600"/>
 
 [[autodoc]] GotOcr2Processor
 
+## GotOcr2Model
+
+[[autodoc]] GotOcr2Model
+
 ## GotOcr2ForConditionalGeneration
 
 [[autodoc]] GotOcr2ForConditionalGeneration

docs/source/en/model_doc/instructblip.md

@@ -69,6 +69,10 @@ The attributes can be obtained from model config, as `model.config.num_query_tok
 [[autodoc]] InstructBlipQFormerModel
     - forward
 
+## InstructBlipModel
+
+[[autodoc]] InstructBlipModel
+
 ## InstructBlipForConditionalGeneration
 
 [[autodoc]] InstructBlipForConditionalGeneration

docs/source/en/model_doc/instructblipvideo.md

@@ -73,6 +73,10 @@ The attributes can be obtained from model config, as `model.config.num_query_tok
 [[autodoc]] InstructBlipVideoQFormerModel
     - forward
 
+## InstructBlipVideoModel
+[[autodoc]] InstructBlipVideoModel
+    - forward
+
 ## InstructBlipVideoForConditionalGeneration
 
 [[autodoc]] InstructBlipVideoForConditionalGeneration

docs/source/en/model_doc/internvl.md

@@ -340,6 +340,11 @@ This example showcases how to handle a batch of chat conversations with interlea
 [[autodoc]] InternVLVisionModel
     - forward
 
+## InternVLModel
+
+[[autodoc]] InternVLModel
+    - forward
+
 ## InternVLForConditionalGeneration
 
 [[autodoc]] InternVLForConditionalGeneration

docs/source/en/model_doc/llava.md

@@ -256,6 +256,10 @@ A list of official Hugging Face and community (indicated by 🌎) resources to h
 
 [[autodoc]] LlavaProcessor
 
+## LlavaModel
+
+[[autodoc]] LlavaModel
+
 ## LlavaForConditionalGeneration
 
 [[autodoc]] LlavaForConditionalGeneration

docs/source/en/model_doc/llava_next.md

@@ -315,6 +315,10 @@ model = AutoModelForImageTextToText.from_pretrained(
 
 [[autodoc]] LlavaNextProcessor
 
+## LlavaNextModel
+
+[[autodoc]] LlavaNextModel
+
 ## LlavaNextForConditionalGeneration
 
 [[autodoc]] LlavaNextForConditionalGeneration

docs/source/en/model_doc/llava_next_video.md

@@ -262,6 +262,10 @@ model = LlavaNextVideoForConditionalGeneration.from_pretrained(
 
 [[autodoc]] LlavaNextVideoImageProcessor
 
+## LlavaNextVideoModel
+
+[[autodoc]] LlavaNextVideoModel
+
 ## LlavaNextVideoForConditionalGeneration
 
 [[autodoc]] LlavaNextVideoForConditionalGeneration

docs/source/en/model_doc/llava_onevision.md

@@ -313,6 +313,10 @@ model = LlavaOnevisionForConditionalGeneration.from_pretrained(
 
 [[autodoc]] LlavaOnevisionVideoProcessor
 
+## LlavaOnevisionModel
+
+[[autodoc]] LlavaOnevisionModel
+
 ## LlavaOnevisionForConditionalGeneration
 
 [[autodoc]] LlavaOnevisionForConditionalGeneration

docs/source/en/model_doc/mistral3.md

@@ -227,6 +227,9 @@ This example also how to use `BitsAndBytes` to load the model in 4bit quantizati
 
 [[autodoc]] Mistral3Config
 
+## Mistral3Model
+
+[[autodoc]] Mistral3Model
 
 ## Mistral3ForConditionalGeneration
 

docs/source/en/model_doc/mllama.md

@@ -130,6 +130,10 @@ print(processor.decode(output[0], skip_special_tokens=True))
 [[autodoc]] MllamaTextModel
     - forward
 
+## MllamaModel
+
+[[autodoc]] MllamaModel
+
 ## MllamaForCausalLM
 
 [[autodoc]] MllamaForCausalLM

docs/source/en/model_doc/paligemma.md

@@ -174,6 +174,10 @@ visualizer("<img> What is in this image?")
 
 [[autodoc]] PaliGemmaProcessor
 
+## PaliGemmaModel
+
+[[autodoc]] PaliGemmaModel
+
 ## PaliGemmaForConditionalGeneration
 
 [[autodoc]] PaliGemmaForConditionalGeneration

docs/source/en/model_doc/qwen2_5_vl.md

@@ -240,6 +240,10 @@ model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
 
 [[autodoc]] Qwen2_5_VLProcessor
 
+## Qwen2_5_VLTextModel
+
+[[autodoc]] Qwen2_5_VLTextModel
+    - forward
 
 ## Qwen2_5_VLModel
 

docs/source/en/model_doc/qwen2_vl.md

@@ -296,6 +296,11 @@ model = Qwen2VLForConditionalGeneration.from_pretrained(
 
 [[autodoc]] Qwen2VLProcessor
 
+## Qwen2VLTextModel
+
+[[autodoc]] Qwen2VLTextModel
+    - forward
+    
 ## Qwen2VLModel
 
 [[autodoc]] Qwen2VLModel

docs/source/en/model_doc/video_llava.md

@@ -215,6 +215,10 @@ model = VideoLlavaForConditionalGeneration.from_pretrained(
 
 [[autodoc]] VideoLlavaProcessor
 
+## VideoLlavaModel
+
+[[autodoc]] VideoLlavaModel
+
 ## VideoLlavaForConditionalGeneration
 
 [[autodoc]] VideoLlavaForConditionalGeneration

docs/source/en/model_doc/vipllava.md

@@ -101,6 +101,10 @@ A chat between a curious human and an artificial intelligence assistant. The ass
 
 [[autodoc]] VipLlavaConfig
 
+## VipLlavaModel
+
+[[autodoc]] VipLlavaModel
+
 ## VipLlavaForConditionalGeneration
 
 [[autodoc]] VipLlavaForConditionalGeneration