I have the following 2 lines of code
x_packed = nn.utils.rnn.pack_padded_sequence(x[sorted_index_th], lengths_sorted, batch_first=True)
print(nn.utils.rnn.pad_packed_sequence(x_packed, batch_first=True)[0].shape)
and I got the following outputs
torch.Size([39, 377, 300])
torch.Size([39, 121, 300])
I wonder why this happens?
sorted_index_th is a 1 dimensional array of indices to change the order of the first dimension of x.