NVIDIA and Baker Lab release RFD4-Proteína for Diffusers
AINVIDIA and the Baker Lab release RFD4-Proteína for Diffusers, a flow-matching model that generates protein sequences and all-atom structures together, conditioned on a binding target, enzyme motif, ligand, or structural features. The repository has 435.6 million parameters in FP32 (1.74 GB) that run in BF16, and it requires unreleased Diffusers support from pull request #15088 plus the matching rfd4-proteina-prep package.
AIWhy it matters.
The release shows how a joint sequence and structure design model runs in Diffusers, with inputs, outputs, and precision and speed tradeoffs for practitioners to weigh.
Source details
NVIDIA · new models on Hugging Face · huggingface.coPublished · added here
Original title: nvidia/RFD4-Proteina-Diffusers