What's new in Diffusers? 🎨

Hugging Face Blog · 3 years ago

Hugging Face released version 0.3 of its Diffusers library with new features for text-to-image generation including image-to-image pipelines, textual inversion, and inpainting capabilities. Stable Diffusion now runs on consumer GPUs with as little as 3.2GB of VRAM through optimizations and on Apple M1/M2 chips via PyTorch's Metal Performance Shaders. The library enables broader access to diffusion models across different hardware platforms and allows users to customize models with their own images in just 3-5 training samples.