Model card: 2026-10-19 release, Diffusers branch

README.md

@@ -29,16 +29,17 @@ This repository is the [Diffusers](https://github.com/huggingface/diffusers) rel
 
 > ### This release requires unreleased Diffusers support
 >
-> These weights are the **2026-10-15 unified release**: a 147-class residue vocabulary and a
-> 32-dimensional latent, with `tsr_matched_noise` and the refreshed sampler recipe. Support for
-> them is not in a released Diffusers version yet — it lives on branch `rfd4-unified-checkpoint`
-> (`0291146a`) of `huggingface/diffusers-new-model-addition-RFD4-Proteina`. An older Diffusers
-> will fail to load this repository rather than sample it incorrectly.
+> These weights are the **2026-10-19 release**: a 147-class residue vocabulary and a 32-dimensional
+> latent, produced by a LoRA fine-tune of the pretrained base that is merged into these weights.
+> Support for them is not in a released Diffusers version yet. It lives on branch
+> `add-rfd4-proteina-torch` of `huggingface/diffusers-new-model-addition-RFD4-Proteina`
+> (pull request 4). An older Diffusers will fail to load this repository rather than sample it
+> incorrectly.
 >
 > You also need the matching [`nvidia/RFD4-Proteina-prep`](https://huggingface.co/nvidia/RFD4-Proteina-prep),
 > which supplies the chirality and element features this checkpoint's encoder requires.
 >
-> **The `examples/*/design.cif` outputs were generated with the previous release** and are kept
+> **The `examples/*/design.cif` outputs were generated with an earlier release** and are kept
 > for reference. They will not reproduce under these weights: the model, the vocabulary and the
 > sampler recipe all changed. The `input.cif` files remain valid.