Differences with torch.nn.LeakyReLU

torch.nn.LeakyReLU

class torch.nn.LeakyReLU(negative_slope=0.01, inplace=False)(input) -> Tensor

For more information, see torch.nn.LeakyReLU.

mindspore.nn.LeakyReLU

class mindspore.nn.LeakyReLU(alpha=0.2)(x) -> Tensor

For more information, see mindspore.nn.LeakyReLU.

Differences

PyTorch: Implements the basic functions of the Leaky ReLU activation function, where the parameter negative_slope is used to control the slope of the activation function, and the parameter inplace is used to control whether to choose to perform the operation of the activation function in-place.

MindSpore: MindSpore API basically implements the same function as PyTorch, where the parameter alpha is the same as the parameter negative_slope in PyTorch, with different parameter names and different default values. However, MindSpore does not have the inplace parameter.

Categories	Subcategories	PyTorch	MindSpore	Difference
Parameters	Parameter 1	negative_slope	alpha	Same function, different parameter names, different default values
	Parameter 2	inplace	-	This parameter is used in PyTorch to control whether to choose to perform the activation function in-place. MindSpore does not have this parameter
Input	Single input	input	x	Same function, different parameter names

Code Example 1

PyTorch parameter negative_slope and MindSpore parameter alpha have the same function, with different parameter names and different default values, and get the same result when both values are the same.

# PyTorch
import torch
import torch.nn as nn

m = nn.LeakyReLU(0.2)
inputs = torch.tensor([[-1.0, 4.0, -8.0], [2.0, -5.0, 9.0]], dtype=float)
output = m(inputs).to(torch.float32).detach().numpy()
print(output)
# [[-0.2  4.  -1.6]
#  [ 2.  -1.   9. ]]

# MindSpore
import mindspore
from mindspore import Tensor
import mindspore.nn as nn

x = Tensor([[-1.0, 4.0, -8.0], [2.0, -5.0, 9.0]]).astype('float32')
m = nn.LeakyReLU()
output = m(x)
print(output)
# [[-0.2  4.  -1.6]
#  [ 2.  -1.   9. ]]

Code Example 2

PyTorch parameter inplace is used to control whether to perform the activation function operation in-place, that is, to perform the activation function operation directly on the input, where the input is changed. MindSpore does not have this parameter, but can assign the output to the input to achieve similar function.

# PyTorch
import torch
import torch.nn as nn

m = nn.LeakyReLU(0.2, inplace=True)
inputs = torch.tensor([[-1.0, 4.0, -8.0], [2.0, -5.0, 9.0]], dtype=torch.float32)
output = m(inputs)
print(inputs.detach().numpy())
# [[-0.2  4.  -1.6]
#  [ 2.  -1.   9. ]]

# MindSpore
import mindspore
from mindspore import Tensor
import mindspore.nn as nn

x = Tensor([[-1.0, 4.0, -8.0], [2.0, -5.0, 9.0]]).astype('float32')
m = nn.LeakyReLU()
x = m(x)
print(x)
# [[-0.2  4.  -1.6]
#  [ 2.  -1.   9. ]]