Views
No views yet
1@article{hoarfrost2022deep,
2 title={Deep learning of a bacterial and archaeal universal language of life
3 enables transfer learning and illuminates microbial dark matter},
4 author={Hoarfrost, Adrienne and Aptekmann, Ariel and Farfanuk, Gaetan and Bromberg, Yana},
5 journal={Nature Communications},
6 volume={13},
7 number={1},
8 pages={2606},
9 year={2022},
10 publisher={Nature Publishing Group}
11}| Architecture | LookingGlass encoder + classification head |
| Encoder | AWD-LSTM (3-layer, unidirectional) |
| Classes | 3 classes: psychrophilic, mesophilic, thermophilic |
| Parameters | ~17M |
1pip install torch
2git clone https://huggingface.co/HoarfrostLab/LGv1_OptimalTempClassifier
3cd LGv1_OptimalTempClassifier1from lookingglass_classifier import LookingGlassClassifier, LookingGlassTokenizer
2
3model = LookingGlassClassifier.from_pretrained('.')
4tokenizer = LookingGlassTokenizer()
5model.eval()
6
7inputs = tokenizer(["GATTACA", "ATCGATCGATCG"], return_tensors=True)
8
9# Get predictions
10predictions = model.predict(inputs['input_ids'])
11print(predictions) # tensor([class_idx, class_idx])
12
13# Get probabilities
14probs = model.predict_proba(inputs['input_ids'])
15print(probs.shape) # torch.Size([2, 3])
16
17# Get raw logits
18logits = model(inputs['input_ids'])
19print(logits.shape) # torch.Size([2, 3])