Kandinsky 2.2 Decoder
Parameters
1.22B
Modality
Image
Architecture
Diffusion Transformer
License
Apache 2.0
Release Date
12 Jul 2023
API Pricing (per 1M)
Self-hosted only
Hardware Sizing
Precision
Workload Memory: 3.11 GB
INT4 · 1024x1024
DiT Backbone Weights: 0.57 GB
VAE Decode Workspace: 0.05 GB
Runtime & Graphs: 2.50 GB
Minimum Required Hardware
1x NVIDIA L4 24GB
3.1 / 21.6 GB usable
1x RTX 3060 12GB
3.1 / 10.8 GB usable
Apple M4 24GB
3.1 / 18.0 GB usable
Determine requirements for your specific workload:
Open in VRAM Calculator
Architecture Diagram
About Kandinsky 2.2 Decoder
Kandinsky 2.2 Decoder is a 1.2B UNet diffusion model combining CLIP ViT-G/14 image embeddings and text conditioning for versatile text-to-image synthesis and image variations.
Technical Specifications
Attention
Attention Heads
12
Attention Head Dimension
64
Dimensions
Hidden Dimension Size
768
Number of Layers
12