Function Differences with tf.nn.max_pool2d

tf.nn.max_pool2d

tf.nn.max_pool2d(
    input,
    ksize,
    strides,
    padding,
    data_format='NHWC',
    name=None
) -> Tensor

For more information, see tf.nn.max_pool2d.

mindspore.nn.MaxPool2d

class mindspore.nn.MaxPool2d(
    kernel_size=1,
    stride=1,
    pad_mode='valid',
    data_format='NCHW'
)(x) -> Tensor

For more information, see mindspore.nn.MaxPool2d.

Differences

TensorFlow: Perform two-dimensional maximum pooling operations on the input multidimensional data.

MindSpore: MindSpore API basically implements the same function as TensorFlow.

Categories	Subcategories	TensorFlow	MindSpore	Differences
Parameters	Parameter 1	input	x	Same function, different parameter names
	Parameter 2	ksize	kernel_size	Same function, different parameter names, no default values for TensorFlow
	Parameter 3	strides	stride	Same function, different parameter names, no default values for TensorFlow
	Parameter 4	padding	pad_mode	Same function, different parameter names, no default values for TensorFlow. For more information, see Conv and Pooling
	Parameter 5	data_format	data_format	-
	Parameter 6	name	-	Not involved

Code Example 1

In TensorFlow, when padding=”SAME”, corresponding to MindSpore with pad_mode=”same” and data_format=”NHWC”, and then set ksize=3 and strides=2 to perform the maximum pooling operation on the input data in two dimensions. The two APIs achieve the same function.

# TensorFlow
import tensorflow as tf

x = tf.constant([[[[1, 2, 3, 4, 5, 6, 7, 8, 9, 10], [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]]]], dtype=tf.float32)
output = tf.nn.max_pool2d(x, ksize=3, strides=2, padding="SAME")
print(output.shape)
# (1, 1, 1, 10)

# MindSpore
import mindspore
import numpy as np
from mindspore import Tensor

device = mindspore.get_context("device_target")
x = Tensor(np.array([[[[1, 2, 3, 4, 5, 6, 7, 8, 9, 10], [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]]]]).astype(np.float32))
if device == "Ascend" or device == "CPU":
    max_pool = mindspore.nn.MaxPool2d(kernel_size=3, stride=2, pad_mode='same')
    x = mindspore.ops.transpose(x, (0, 3, 2, 1))
    output = max_pool(mindspore.Tensor(x))
    output = mindspore.ops.transpose(output, (0, 3, 2, 1))
    print(output.shape)
# (1, 1, 1, 10)
else:
    max_pool = mindspore.nn.MaxPool2d(kernel_size=3, stride=2, pad_mode='same', data_format='NHWC')
    output = max_pool(x)
    print(output.shape)
# (1, 1, 1, 10)