Numpy

If you want to make batches, where N+1 batch will follow N batch, try this:

import numpy as np

batch_quantity = 3
batch_size = 2
some_embedding = 4

initial_data = np.arange(24)
reshaped = initial_data.reshape(batch_size, batch_quantity, some_embedding)
result_data = reshaped.transpose(1, 0, 2)

print(result_data)
# [[[ 0  1  2  3]
#   [12 13 14 15]]
#
#  [[ 4  5  6  7]
#   [16 17 18 19]]
#
#  [[ 8  9 10 11]
#   [20 21 22 23]]]

Another info notes:

Change display methods

np.set_printoptions(precision=2)
np.set_printoptions(suppress=True)

Split some array in folds(numpy)

y_folds = np.array_split(y_digits, 3)

Shuffle the indices for test data

indexes = np.random.permutation(len(iris_X))

Select max value based on another array max value(numpy)

best_alpha = alphas[scores.index(max(scores))]

Reshape one array to another one shape

face_compressed.shape = face.shape

Reshape numpy array with dynamically calculated second dimension

test_data = np.zeros((4, 3))
test_data.shape  # (4, 3)
reshaped = test_data.reshape((6, -1))
reshaped.shape  # (6, 2)

Create list of 3 random items form 0 to 255

np.random.randint(0, high=256, size=(3, )).tolist()

Assign values to one line in array(notes for tensorflow) .. code-block:: python

M_t = np.arange(15).reshape(5, 3) # array([[ 0, 1, 2], # [ 3, 4, 5], # [ 6, 7, 8], # [ 9, 10, 11], # [12, 13, 14]]) indexes = np.array([0, 1, 0, 0, 0]) new_value = np.arange([101, 102, 103]) (M_t.T * (-1 * (indexes -1))).T + np.outer(indexes, new_value) # array([[ 0, 1, 2], # [103, 104, 105], # [ 6, 7, 8], # [ 9, 10, 11], # [ 12, 13, 14]])