CLIPVariant

Enums · AiDotNet.Enums

Variants for CLIP text encoder models.

Fields

Field Summary
CLIPVariant StableDiffusionX4Upscaler Frozen OpenCLIP ViT-H text tower shipped with the Stable Diffusion x4 Upscaler: 1024 hidden dimensions, 23 layers, 16 heads, and a 512-dim pooled projection.
CLIPVariant ViTBigG14 CLIP ViT-bigG/14: 1280-dim, 32 layers.
CLIPVariant ViTH14 CLIP ViT-H/14: 1024-dim, 24 layers.
CLIPVariant ViTL14 CLIP ViT-L/14: 768-dim, 12 layers.
int value__