AI updates
Modular Diffusers: diffusion pipelines from composable blocks
Hugging Face Diffusers introduced ModularPipeline, which builds diffusion pipelines from reusable blocks (text encoding, image encoding, denoising, decoding) that can be inspected, swapped and run on their own.
- A pipeline becomes a list of blocks (text encoder, VAE encoder, denoise, decode) you can inspect, reorder or replace.
- Defining the workflow and loading the weights are separate steps.