[Docs] Reduce time spent generating API docs (#34255)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

vllm/model_executor/models/blip2.py

@@ -73,6 +73,7 @@ class Blip2ImageEmbeddingInputs(TensorSchema):
 
 
 Blip2ImageInputs: TypeAlias = Blip2ImagePixelInputs | Blip2ImageEmbeddingInputs
+"""Alias for supported BLIP-2 image input types."""
 
 
 class Blip2QFormerMultiHeadAttention(nn.Module):

vllm/model_executor/models/llava.py

@@ -121,6 +121,7 @@ class LlavaImageEmbeddingInputs(TensorSchema):
 LlavaImageInputs: TypeAlias = (
     LlavaImagePixelInputs | PixtralHFImagePixelInputs | LlavaImageEmbeddingInputs
 )
+"""Alias for supported LLaVA image input types."""
 
 
 class LlavaMultiModalProjector(nn.Module):

vllm/model_executor/models/llava_next.py

@@ -78,6 +78,7 @@ class LlavaNextImageEmbeddingInputs(TensorSchema):
 LlavaNextImageInputs: TypeAlias = (
     LlavaNextImagePixelInputs | LlavaNextImageEmbeddingInputs
 )
+"""Alias for supported LLaVA-NeXT image input types."""
 
 
 class LlavaNextLikeConfig(LlavaLikeConfig, Protocol):
@@ -106,6 +107,7 @@ class LlavaNextProcessingInfo(BaseLlavaProcessingInfo):
         image_width: int,
         image_height: int,
     ) -> int:
+        """Get the number of image tokens for the given image dimensions."""
         hf_config = self.get_hf_config()
         vision_encoder_info = self.get_vision_encoder_info()