# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForMaskedLM
tokenizer = AutoTokenizer.from_pretrained("nilsho01/LittleNano")
model = AutoModelForMaskedLM.from_pretrained("nilsho01/LittleNano", device_map="auto")Quick Links
General idea
This model was trained on about 9 million nanobody sequences. For that, the full sequence length was parsed to the model. Thus, it may be recommended to use the pad token if a region such as CDR3 should be analysed isolated. The model performed better in qualitative analysis in comparison to the ESM2 model of the same size equally well to existing models with imunological knowledge like Antiberta2. The strength of the model is that it is a small language model and will be very fast for downstream tasks which makes it interesting for industrial use cases.
- Downloads last month
- 23
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("token-classification", model="nilsho01/LittleNano")