Tensors in PyTorch can have non-contiguous memory, as described here.
What about Tensorflow? Can tensors in Tensorflow have non-contiguous memory? Or is it impossible because of static graph?
Asked
Active
Viewed 180 times
