I am new to pytorch I have a for loop running, where a tensor (of which the size is unknown) is being returned. I need some way out to store the tensor returned in a temporary variable and use it to compare with the tensor in the next loop. But I dont know how to initialize a tensor of unknown size.
I may have misunderstood your needs, but Iād just stick it in a python variable as you normally would. e.g.
for stuff in some_list:
output = do_stuff_with(stuff)
# compare temp and out
temp = output
tensor = torch.Tensor()
gives you an empty tensor. You can resize this with tensor.resize_as_(output)
when you have another tensor.
1 Like