Function Differences with tf.compat.v1.assign_sub

tf.compat.v1.assign_sub

tf.compat.v1.assign_sub(ref, value, use_locking=None, name=None) -> Tensor

For more information, see tf.compat.v1.assign_sub.

mindspore.ops.assign_sub

mindspore.ops.assign_sub(variable, value)-> Tensor

For more information, see mindspore.ops.assign_sub.

Differences

TensorFlow: Update the network parameters by subtracting a specific value from the network parameters, and return a Tensor with the same type as ref.

MindSpore: MindSpore API implements the same functions as TensorFlow, with some different parameter names.

Categories	Subcategories	TensorFlow	MindSpore	Differences
Parameters	Parameter 1	ref	variable	Same function, different parameter names
	Parameter 2	value	value	-
	Parameter 3	use_locking	-	In TensorFlow, whether to use locks in update operations. Default value: False.
	Parameter 4	name	-	Not involved

Code Example 1

The outputs of MindSpore and TensorFlow are consistent.

# TensorFlow
import tensorflow as tf
import numpy as np

variable = tf.Variable(np.array([[2.4, 1], [0.1, 6]]), dtype=tf.float32)
value = tf.constant(np.array([[-2, 3], [3.6, 1]]), dtype=tf.float32)
out = tf.compat.v1.assign_sub(variable, value)
print(out.numpy())
# [[ 4.4 -2. ]
#  [-3.5  5. ]]

# MindSpore
import mindspore
import numpy as np
from mindspore.ops import function as ops
from mindspore import Tensor

variable = Tensor(np.array([[2.4, 1], [0.1, 6]]), mindspore.float32)
value = Tensor(np.array([[-2, 3], [3.6, 1]]), mindspore.float32)
out = ops.assign_sub(variable, value)
print(out)
# [[ 4.4 -2. ]
#  [-3.5  5. ]]