Views
No views yet
640 (fp16).embed_dim=640, encoder_depth=28, encoder_heads=20.150M.ckpt: full Lightning checkpoint.config.yaml: model + data config. The path fields are placeholders, point them at your local data.backbone_tokenizer.pt, fullatom_tokenizer.pt: structure tokenizers (PDB to token IDs). See Acknowledgements.1pip install torch huggingface_hub omegaconf numpy lmdb biotite
2git clone https://github.com/hsjang0/TriProRep.git
3cd TriProRep1import sys; sys.path.insert(0, "code/triprorep")
2from inference import load_encoder, embed_pdb
3
4encoder = load_encoder("150M", hf_repo="k-fold-structure/triprorep-150M")
5features = embed_pdb(encoder, "your_protein.pdb",
6 hf_repo="k-fold-structure/triprorep-150M")
7print(features.shape) # (L, 640) fp16embed_pdb downloads the bundled tokenizers from this repo on first call,
then runs PDB to (seq, bb, fa) tokens to encoder. If you already have token
IDs (e.g. from k-fold-structure/repsp-triprorep-tokens), call
encode(encoder, seq, bb, fa) directly. For CPU, pass device="cpu" to
load_encoder.backbone_tokenizer.pt (aminoaseed VQ-VAE) is from
StructTokenBench.1@misc{triprorep,
2 title = {Atom-level Protein Representation Learning Improves Protein Structure Prediction},
3 author = {Kim, Taewon and Jang, Hyosoon and Seo, Hyunjin and Seo, Seonghwan and Kim, Hyeongwoo and Zhung, Wonho and Shin, Mingyeong and Kim, Wooyoun and Ahn, Sungsoo},
4 year = {2026},
5 eprint = {2605.22133},
6 archivePrefix = {arXiv},
7 primaryClass = {cs.LG},
8 url = {https://arxiv.org/abs/2605.22133}
9}