miae-b / README.md
nielsr's picture
nielsr HF Staff
Improve model card and update metadata
65edee0 verified
|
raw
history blame
2.12 kB
metadata
library_name: tedbench
license: bsd-3-clause
pipeline_tag: graph-ml
tags:
  - protein
  - structure
  - fold-classification
  - tedbench

TEDBench — Pretrained autoencoder (structure only)

Variant: miae_b  |  Parameters: 102M  |  Layers: 12  |  Hidden dim: 768  |  Attn heads: 12

This repository contains the pretrained MiAE-B (base) checkpoint, a self-supervised model for protein structure representation learning introduced in the paper Protein Fold Classification at Scale: Benchmarking and Pretraining.

MiAE (Masked Invariant Autoencoders) is an $\mathrm{SE(3)}$-invariant autoencoder that uses an extremely high masking ratio (up to 90%) of backbone frames to reconstruct full protein structures. It can be used as a feature extractor or as the starting point for fine-tuning on protein fold classification tasks.

  • Authors: Dexiong Chen, Andrei Manolache, Mathias Niepert, Karsten Borgwardt
  • Official Code: BorgwardtLab/TEDBench

Architecture sizes

Variant Params Layers Hidden dim Attn heads
miae_s 29 M 6 512 8
miae_b 102 M 12 768 12
miae_l 339 M 24 1 024 16

Append +model.use_seq_input=true to miae_b for the +seq variant.

Usage

Load from the HuggingFace Hub

Using the tedbench library:

from tedbench.utils.io import load_from_hf

model = load_from_hf("TEDBench/miae-b")
model.eval()

From a Lightning checkpoint

from tedbench.model import MiAE

model = MiAE.load_from_checkpoint("model.ckpt", weights_only=False)
model.eval()

Citation

@inproceedings{chen2026tedbench,
  title={Protein Fold Classification at Scale: Benchmarking and Pretraining},
  author={Chen, Dexiong explorer and Manolache, Andrei and Niepert, Mathias and Borgwardt, Karsten},
  booktitle={Proceedings of the 43rd International Conference on Machine Learning (ICML)},
  year={2026}
}