I know in tensorflow I can do it with tf.slice(my_tensor, begin, size)
Many thanks
In Torch / PyTorch you can use the narrow(dimension, start, length)
method (ref.), which allows you to get a smaller range (length
) of a given dimension
at a specific start
ing point (so, you’ll end up with the same number of dimensions).
Use instead the select(dimension, index)
method (ref.) to choose a specific index
of one dimension
(you’ll end up with one less number of dimensions).
2 Likes
Or better just use the regular indexing syntax, as if you were operating on numpy arrays: arr[0, 2:4]
will select first slice along the first dimension and slices 2 and 3 from the second dimension.
7 Likes