| Model | Dataset | Description |
|---|---|---|
| AbLingua-300M | OAS-Unpaired-300M | A small language model that is easy to fine-tune |
| AbLingua-600M | OAS-Unpaired-300M | Medium-sized model that can support a variety of downstream tasks and performs well in antibody structure prediction |
As a prerequisite, you must have PyTorch and Transformers installed to use this repository. PyTorch and Transformers only require the latest version to be installed.
# Python >= 3.10
pip install torch
pip install transformers# weight should in weight floder
cd AbLingua/weight/
# download model json
wget -c 'https://huggingface.co/IDEA-XtalPi/AbLingua/resolve/main/config.json'
# download model weight
wget -c 'https://huggingface.co/IDEA-XtalPi/AbLingua/resolve/main/pytorch_model.bin'Representations from AbLinuga may be useful as features for deep learning models.
from AbLingua.embed import get_collator, get_model
collator = get_collator()
model = get_model()
seq = ['QVTLRESGPAL',
'VKPTQTLTLTC']
seq_input = collator(seq)
tokens_embedding = model(**seq_input).hidden_states[-1]
# tokens_embedding.shape
# [2, 256, 1280]