We may find three dots (…) or ellipsis when get numpy ndarray elements. For example:
shift_logits = logits[..., :-1, :] shift_labels = labels[..., 1:]
What this ellipsis means? In this tutorial, we will use an example to show you.
For example:
import numpy as np
x = np.arange(0,32).reshape(2, 4, 4)
print("x =")
print(x)
print("x[:,:, 1] = ")
print(x[:,:, 1])
print("x[..., 1] = ")
print(x[..., 1])Run this code, we will see this result.
x = [[[ 0 1 2 3] [ 4 5 6 7] [ 8 9 10 11] [12 13 14 15]] [[16 17 18 19] [20 21 22 23] [24 25 26 27] [28 29 30 31]]] x[:,:, 1] = [[ 1 5 9 13] [17 21 25 29]] x[..., 1] = [[ 1 5 9 13] [17 21 25 29]]
From the result above, we can find:
x[:,:, 1] == x[…, 1]
Actually, ellipsis mean :,:,…,: from the beginning and :,:,…,: from the ending.
For example:
if the shape of x is [3,4,5,6,]
x[:,:,2:3,:] = […,2:3,…]