INNER CODE UNIT · Python

predict

graviraja/100-Days-of-NLP · applications/classification/toxic_comment_classification/app/app.py:47

def predict(sentence, bpe_model, model):
    model.eval()

    if isinstance(sentence, str):
        sentence = preprocess(sentence)
        tokens = bpe_tokenizer(sentence)
    else:
        tokens = [int(token) for token in sentence]

    src_indexes = tokens
 
    # convert to tensor format
    # since the inference is done on single sentence, batch size is 1
    src_tensor = torch.LongTensor(src_indexes).unsqueeze(1).to(device)
    # src_tensor => [seq_len, 1]

    src_length = torch.LongTensor([len(src_indexes)])
    # src_length => [1]

View source record →