3d054be72e81af7acb1109028d7be4fceebae720,train.py,,test,#Any#Any#Any#,58

Before Change

            S2 = S2.cuda()
            labels = labels.cuda()
        // Wrap to autograd.Variable
        X, S1, S2 = Variable(X), Variable(S1), Variable(S2)
        // Forward pass
        outputs, predictions = net(X, S1, S2, config)
        // Select actions with max scores(logits)

After Change

        if X.size()[0] != config.batch_size:
            continue  // Drop those data, if not enough for a batch
        // automaticlly select device, device agnostic 
        device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
        X = X.to(device)
        S1 = S1.to(device)
        S2 = S2.to(device)

Instances