Created by: LogicZMaksimka
Few-Shot Text Classification Model
-
Added few-shot text classification model based on DNNC architecture that does not require any training. It takes an example that needs to be classified, a dataset based on which classification is performed and predicts the label of the example.
-
At the moment, main config supports only predictions. Training and evaluation have to be conducted in separate configuration files.
-
Model supports out-of-scope detection by using a confidence threshold