Diffusers adds LoRA fine-tuning for NVIDIA's RFD4-Proteina protein design model

ArashVahdat · x · 2026-11-21

Hugging Face announces that Diffusers now supports LoRA fine-tuning of NVIDIA's RFD4-Proteina protein design model. The example builds a dataset with atomworks and trains rank-32 adapters on the denoiser with flow-matching losses computed on backbone coordinates and local latents instead of pixels. The training code closely mirrors standard image/video generation workflows, lowering the bar for domain experts.

Related event: Diffusers Adds LoRA Fine-Tuning Support for NVIDIA RFD4-Proteina(2 posts)→

Original post →

More from coding & agent

coding & agent channel →