Rebuilt from the current root bundle (weights 5b549dd6), replacing a conversion of the superseded 5f00cdcd.
Also follows the release on three settings the earlier conversion predated: the UNK residue's atom37 row now carries its backbone, thinking_medium is the front-loaded 180-pass schedule rather than the earlier peaked 185, and recycle_remaining is one class wide, which the native config now leaves implicit.
Parity against the native runtime: feature factories and autoencoder exact on monomer, ligand and glycan inputs, denoiser within 2.3e-3.
3 days ago
Rename projections to the transformers convention
Regenerated from the same release bundle: every tensor is byte-identical, only the keys change (to_q -> q_proj, to_out_node -> o_proj, to_gamma -> scale_proj, ...), matching the renamed modelling code.
The config drops nine flags the released architecture pins and folds the retired sc_scale_score/tsr_matched_noise sampler controls into sc_scale_noise=1.0, which describes the same trajectory.
4 days ago
Add the Transformers checkpoint for the 2026-10-15 release
Converted from this repository's native bundle (weights_sha256 5f00cdcd...) with the unified-vocabulary converter: 147-class residue vocabulary, latent dim 32, fused query/key/value projections split. The sampler recipe is pinned from main's composed inference config (thinking_medium, 60 steps, 185 trunk passes, tsr_matched_noise, t_lim_ode 0.95, tsr_sigma 12.0) rather than taken from converter defaults. Lives in a subfolder because a Transformers export writes config.json and model.safetensors, the names the native bundle already uses at the root.
