Python Deep Learning Projects

Training the seq2seq model

Now, let's train the model. We will need some helper functions for the padding of the sentence and to calculate the accuracy of the model:

def pad_sentence_batch(sentence_batch, pad_int):    padded_seqs = []    seq_lens = []    max_sentence_len = 50    for sentence in sentence_batch:        padded_seqs.append(sentence + [pad_int] * (max_sentence_len - len(sentence)))        seq_lens.append(50)    return padded_seqs, seq_lensdef check_accuracy(logits, Y):    acc = 0    for i in range(logits.shape[0]):        internal_acc = 0        for k in range(len(Y[i])):            if Y[i][k] == logits[i][k]:                internal_acc += 1        acc += (internal_acc / len(Y[i]))    return acc / logits.shape[0]

We initialize our model and iterate the session for the defined number of epochs:

tf.reset_default_graph() ...