Home

Looking for a Guide?

Check out the ELI5 Training guide.

About

This is a WIP port and extension of Shivam Shriao's Diffusers Repo, which is a modified version of the default Huggingface Diffusers Repo optimized for better performance on lower-VRAM GPUs.

Many new features have been added such as:

• LoRa and LoRa Extended
• Training multiple concepts simultaneously
• Improved dataset processing (e.g. bucketing, caption processing, and caching)
• Expanded optimizer, attention, and scheduler options
• Finer control of class image generation

⚠️Warning⚠️

The main branch is akin to a beta release, we are still squashing all the bugs. Any other branches are early alpha and may break core features. Please be patient and descriptive when creating issues or requesting help in the Discord.