I don't think so. shuffle=True is used to shuffle the order of samples in the training set so the model won't overfit on specific sample orders. But if you have time series data, shuffle=True will mess with the training because the model actually needs to learn the order in this case.