d740663af36c1b392d3beedc4c4c2817365ee49f,ch12/train_model.py,,,#,53

Before Change

            for idx, out_seq in enumerate(out_seq_list):
                r = net.decode_teacher(net.get_encoded_item(enc, idx), out_seq)
                net_results.append(r)
                net_targets.extend(out_idx[idx][1:])
                bleu_sum += model.seq_bleu(r, out_idx[idx][1:])
                bleu_count += 1
            results_v = torch.cat(net_results)
            targets_v = Variable(torch.LongTensor(net_targets))

After Change

            net_targets = []
            for idx, out_seq in enumerate(out_seq_list):
                ref_indices = out_idx[idx][1:]
                if random.random() < TEACHER_PROB:
                    r = net.decode_teacher(net.get_encoded_item(enc, idx), out_seq)
                else:
                    r, _ = net.decode_chain_argmax(embeddings, net.get_encoded_item(enc, idx),
                                                   out_seq.data[0], len(ref_indices))
                net_results.append(r)
                net_targets.extend(ref_indices)
                bleu_sum += model.seq_bleu(r, ref_indices)
                bleu_count += 1

Instances