DFloat11: Lossless LLM Compression for Efficient GPU Inference

[nitinmukesh]

DFloat11 is a lossless compression framework that reduces the size of Large Language Models (LLMs) by approximately 30% while preserving bit-for-bit identical outputs to the original model. It enables efficient GPU inference on resource-constrained hardware without sacrificing accuracy.

Wan and FLUX are supported, if anyone wants to try
https://github.com/LeanModels/DFloat11

Looks interesting and diffusers is supported. I can't try due to VRAM but in case anyone is interested.