transformer_24.py

def predict(test_source_text=None):
    # If test sentence is not provided
    # randomly pick up one from the training data
    if test_source_text is None:
        test_source_text = raw_data_en[np.random.choice(len(raw_data_en))]

    print(test_source_text)
    
    # Tokenize the test sentence to obtain source sequence
    test_source_seq = en_tokenizer.texts_to_sequences([test_source_text])
    print(test_source_seq)

    en_output = encoder(tf.constant(test_source_seq))

    de_input = tf.constant([[fr_tokenizer.word_index['<start>']]], dtype=tf.int64)

    out_words = []

    while True:
        de_output = decoder(de_input, en_output)
        
        # Take the last token as the predicted token
        new_word = tf.expand_dims(tf.argmax(de_output, -1)[:, -1], axis=1)
        out_words.append(fr_tokenizer.index_word[new_word.numpy()[0][0]])

        # The next input is a new sequence
        # contains both the input sequence and the predicted token
        de_input = tf.concat((de_input, new_word), axis=-1)

        # End if hitting <end> or length exceeds 14
        if out_words[-1] == '<end>' or len(out_words) >= 14:
            break

    print(' '.join(out_words))