Module clip 

CLIP image encoder for reference-image conditioning.

Implements the ViT-H/14 CLIP image encoder used to extract per-image embeddings that condition the multi-view diffusion model via IP-adapter cross-attention.

Structs

ClipImageEncoder

CLIP ViT image encoder.

ClipVisionConfig

CLIP vision model configuration (ViT-H/14 defaults).

Functions

build_clip_encoder

Build a CLIP encoder from a DiffusionConfig with default ViT-H/14 settings.