train_ALL_LSTM.py 文件源码

def eval(data_iter, model, args, scheduler):
    model.eval()
    corrects, avg_loss = 0, 0
    for batch in data_iter:
        feature, target = batch.text, batch.label
        target.data.sub_(1)
        # feature.data.t_(), target.data.sub_(1)  # batch first, index align
        # feature.data.t_(),\
        # target.data.sub_(1)  # batch first, index align
        # target = autograd.Variable(target)
        if args.cuda is True:
            feature, target = feature.cuda(), target.cuda()

        model.hidden = model.init_hidden(args.lstm_num_layers, args.batch_size)
        if feature.size(1) != args.batch_size:
            # continue
            model.hidden = model.init_hidden(args.lstm_num_layers, feature.size(1))
        logit = model(feature)
        loss = F.cross_entropy(logit, target, size_average=False)
        # scheduler.step(loss.data[0])

        avg_loss += loss.data[0]
        corrects += (torch.max(logit, 1)[1].view(target.size()).data == target.data).sum()

    size = len(data_iter.dataset)
    avg_loss = loss.data[0]/size
    accuracy = float(corrects)/size * 100.0
    model.train()
    print('\nEvaluation - loss: {:.6f}  acc: {:.4f}%({}/{}) \n'.format(avg_loss,
                                                                       accuracy,
                                                                       corrects,
                                                                       size))