The numpy.zeros_like() method creates a new array of zeros with the same shape and type as a given input array or array-like object.

import numpy as np

arr = np.array([[1, 2, 3],
                [4, 5, 6]])

zeros_array = np.zeros_like(arr)

print(zeros_array)

# Output:
# [[0 0 0]
#  [0 0 0]]

You can see that we create a 2D array of 0s from the input array with the same shape, 2×3, and matching int64 dtype (the default on most systems).

Syntax

numpy.zeros_like(arr, 
                 dtype=None, 
                 order='K', 
                 subok=True, 
                 shape=None)

Parameters

Argument	Description
arr (required, array_like)	It represents an array or array-like object. You can take it as a reference array from which the zeros_like() method will create a new array filled with 0s with the same shape and size as arr.
dtype (data-type, optional)	It overrides the dtype of the resulting array. By default, it is an int64, but you can override it based on your requirements.
order ({‘C’, ‘F’, ‘A’, ‘K’}, optional)	It controls the memory layout of the output array. By default, it is “k”.
subok (bool, optional)	If True (default), they are subclasses of ndarray. If False, the output is always a base ndarray.
shape (int or sequence of ints, optional)	It overrides the shape of the resulting array.

Overriding dtype

Even if the input array contains float values, you can change the output array’s type to int64 by passing the dtype=int argument.

import numpy as np

arr = np.array([1.9, 2.1, 3.5])

zeros_int = np.zeros_like(arr, dtype=int)

print(zeros_int)

# Output: [0 0 0]

Specifying memory order

The ‘order’ control defines the layout; ‘K’ preserves the input’s order for optimal memory access.

import numpy as np

arr = np.array([[1, 2], [3, 4]], order='F')  # Fortran-contiguous

z_c = np.zeros_like(arr, order='C')  # Force C-order

print(z_c)
# Output: [[0 0] [0 0]], C-contiguous

z_k = np.zeros_like(arr)

print(z_k)
# Output: [[0 0] [0 0]], K-contiguous
# Default 'K' preserves 'F'

Empty array input

If the input is an empty array, this method handles empty shapes correctly, producing an empty zero array.

import numpy as np

empty_array = np.array([])  # Shape (0,)

empty_zeros = np.zeros_like(empty_array)

print(empty_zeros)

# Output: []

That’s all!