Calculating accuracy for a multi-label classification problem

vision
2

I came up with the answer: torch.sigmoid(classified_labels).data > 0.5 will give the correct labels with MultiLabelSoftMarginLoss().