20d1da406aaa2ae2a32c48a28e3efe4f475d7398,ch12/train_crossent.py,,,#,18

Before Change

    // initialize embedding lookup table
    embeddings = nn.Embedding(num_embeddings=emb.shape[0], embedding_dim=emb.shape[1])
    embeddings.weight.data.copy_(torch.from_numpy(emb))
    embeddings.weight.requires_grad = False
    if args.cuda:
        embeddings.cuda()

    net = model.PhraseModel(emb_size=emb.shape[1], dict_size=emb.shape[0], hid_size=model.HIDDEN_STATE_SIZE)
    if args.cuda:
        net.cuda()
    log.info("Model: %s", net)

After Change

    // if args.cuda:
    //     embeddings.cuda()

    net = model.PhraseModel(emb_size=model.EMBEDDING_DIM, dict_size=len(emb_dict), hid_size=model.HIDDEN_STATE_SIZE)
    if args.cuda:
        net.cuda()
    log.info("Model: %s", net)

    writer = SummaryWriter(comment="-" + args.name)

    optimiser = optim.Adam(net.parameters(), lr=LEARNING_RATE)
    best_bleu = None
    for epoch in range(MAX_EPOCHES):
        random.shuffle(train_data)
        losses = []
        bleu_sum = 0.0
        bleu_count = 0
        for batch in data.iterate_batches(train_data, BATCH_SIZE):
            optimiser.zero_grad()
            input_seq, out_seq_list, _, out_idx = model.pack_batch(batch, net.emb, cuda=args.cuda)
            enc = net.encode(input_seq)

            net_results = []

Instances