Loading...
Loading...
Browse all stories on DeepNewz
VisitBitsFusion Achieves 7.9X Compression of Stable Diffusion v1.5 to 1.99 Bits with Improved Performance
Jun 7, 2024, 07:20 AM
BitsFusion is a new weight quantization method that compresses the UNet of Stable Diffusion v1.5 from 1.72 GB (FP16) to 219 MB (1.99 bits), achieving a 7.9X compression ratio. This method not only reduces the model size significantly but also improves its performance. The paper introduces a mixed-precision strategy to achieve this compression, transitioning from FP32, making diffusion-based image generation models more efficient and capable of synthesizing high-quality content.
View original story
Markets
Yes • 50%
No • 50%
Official announcements from major AI companies like OpenAI, Google, or Meta
No • 50%
Yes • 50%
Official announcements or software releases from relevant AI platforms
Yes • 50%
No • 50%
Publication in a peer-reviewed journal
Image Generation • 25%
Other • 25%
Natural Language Processing • 25%
Video Compression • 25%
Industry reports, surveys, and official announcements
Above 30% • 25%
0-10% • 25%
11-20% • 25%
21-30% • 25%
Official benchmarks and performance reports
OpenAI • 25%
Google • 25%
Meta • 25%
Other • 25%
Official announcements from AI platforms