How to shuffle variable sequence in PyTorch

nlp
1

I have negative and postive sequences which are divided in two group. How can I combine them to train?

2

Hi,

You can concatenate the two tensors and then do a random shuffle on the combined tensor.

Assuming that the torch tensor containing the positive examples is t1 and that of negative examples is t2, you can do something like this,

t1 = torch.cat((t1, t2), dim=0) # dim=0 assumes that your first axis is the batch axis. data = t1[torch.randperm(t1.size()[0])] # Shuffle the data

Hope this helps,
Cheers!