Clark Air shrinks the Sana image model to 374MB
Clark Air is a heavily compressed version of the main transformer part of the Sana 1.6B text-to-image model. The original FP16 transformer is 3.21GB, while the packed Clark Air file is 374MB, about 8.6 times smaller. The compression uses with group-level scaling values.
About 5% of the model, including conditioning and projection layers, stays at higher precision to reduce quality loss. The measured size is about 1.85 bits per weight. An unpacked version is also available, so it can be swapped into diffusers more easily.
The base model is Sana 1.6B for 512-pixel , and the license is Apache-2.0.
Key points
- Sana 1.6B’s transformer was reduced from 3.21GB to 374MB.
- The packed model is about 8.6 times smaller than the FP16 version.
- It uses while keeping a small important part at higher precision.
- An unpacked version can be used as a drop-in replacement with diffusers.
- The model is released under the .