Function Differences with torch.nn.functional.elu

torch.nn.functional.elu

torch.nn.functional.elu(input, alpha=1.0, inplace=False) -> Tensor

For more information, see torch.nn.functional.elu.

mindspore.ops.elu

mindspore.ops.elu(input_x, alpha=1.0) -> Tensor

For more information, see mindspore.ops.elu.

Differences

PyTorch: compute the exponential linear value of the input x. The result is \( \text{elu}(x) = \max(0,x) + \min(0, \alpha * (\exp(x) - 1)) \), with the inplace parameter to choose whether to operate in-place or not, and the default is False.

MindSpore: MindSpore API implements the same functions as PyTorch, but α currently only supports 1.0.

Categories	Subcategories	PyTorch	MindSpore	Difference
Parameters	Parameter 1	input	input_x	Same function, different parameter names
	Parameter 2	alpha	alpha	α factor. MindSpore currently only supports alpha equal to 1.0
	Parameter 3	inplace	-	MindSpore does not have this parameter

Code Example 1

Both APIs implement the exponential linear unit function, but PyTorch can customize the α coefficient, and MindSpore only supports a coefficient of 1.0.

# PyTorch
import torch
from torch import tensor
from torch.nn.functional import elu
import numpy as np

x_ = np.array([[np.arange(-6,0).reshape(2, 3),np.arange(0,6).reshape(2, 3)]])
x = tensor(x_, dtype=torch.float32)
output = elu(x, alpha = 1).detach().numpy()
print(output)
# [[[[-0.9975212  -0.99326205 -0.9816844 ]
#   [-0.95021296 -0.86466473 -0.63212055]]
#
#  [[ 0.          1.          2.        ]
#   [ 3.          4.          5.        ]]]]

# MindSpore
import mindspore as ms
from mindspore import ops
import numpy as np

x_ = np.array([[np.arange(-6,0).reshape(2, 3),np.arange(0,6).reshape(2, 3)]])
x = ms.Tensor(x_, ms.float32)
output = ops.elu(x)
print(output)
# [[[[-0.9975212  -0.99326205 -0.9816844 ]
#    [-0.95021296 -0.86466473 -0.6321205 ]]
#
#   [[ 0.          1.          2.        ]
#    [ 3.          4.          5.        ]]]]