Hands-On Natural Language Processing with Python

Training

We will now train our network on the French sentence and the corresponding English translation. Before that, we will look into the function that outputs the training batches:

def get_batches(en_text, fr_text, batch_size):    for batch_idx in range(0, len(fr_text)//batch_size):        start_idx = batch_idx * batch_size        en_batch = en_text[start_idx:start_idx + batch_size]        fr_batch = fr_text[start_idx:start_idx + batch_size]        pad_en_batch = np.array(pad_sentences(en_batch, en_word2int))        pad_fr_batch = np.array(pad_sentences(fr_batch,fr_word2int))        pad_en_lens = []        for en_b in pad_en_batch:            pad_en_lens.append(len(en_b))         pad_fr_lens = []        for fr_b in pad_fr_batch:            pad_fr_lens.append(len(fr_b))  yield pad_en_batch, pad_fr_batch, pad_en_lens, pad_fr_lens ...