Function Differences with torch.nn.Softshrink

torch.nn.Softshrink

class torch.nn.Softshrink(lambd=0.5)(input) -> Tensor

For more information, see torch.nn.Softshrink.

mindspore.nn.SoftShrink

class mindspore.nn.SoftShrink(lambd=0.5)(input_x) -> Tensor

For more information, see mindspore.nn.SoftShrink.

Differences

PyTorch: Used to calculate the Softshrink activation function.

MindSpore: The interface name is different from PyTorch. MindSpore is SoftShrink, while PyTorch is Softshrink, and the function is the same.

Categories	Subcategories	PyTorch	MindSpore	Difference
Parameter	Parameter 1	lambd	lambd	-
Input	Single input	input	input_x	Same function, different parameter names

Code Example 1

Compute the SoftShrink activation function for lambd=0.3.

# PyTorch
import numpy as np
import torch
from torch import tensor, nn

m = nn.Softshrink(lambd=0.3)
input_ = np.array([[0.5297, 0.7871, 1.1754], [0.7836, 0.6218, -1.1542]], dtype=np.float32)
input_t = tensor(input_)
output = m(input_t)
print(output.numpy())
# [[ 0.22969997  0.4871      0.8754    ]
#  [ 0.48359996  0.3218     -0.85419995]]

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

m = nn.SoftShrink(lambd=0.3)
input_ = np.array([[0.5297, 0.7871, 1.1754], [0.7836, 0.6218, -1.1542]], dtype=np.float32)
input_t = Tensor(input_, mindspore.float32)
output = m(input_t)
print(output)
# [[ 0.22969997  0.4871      0.8754    ]
#  [ 0.48359996  0.3218     -0.85419995]]

Code Example 2

SoftShrink defaults to lambd=0.5.

# PyTorch
import numpy as np
import torch
from torch import tensor, nn

m = nn.Softshrink()
input_ = np.array([[0.5297, 0.7871, 1.1754], [0.7836, 0.6218, -1.1542]], dtype=np.float32)
input_t = tensor(input_)
output = m(input_t)
print(output.numpy())
# [[ 0.02969998  0.28710002  0.6754    ]
#  [ 0.28359997  0.12180001 -0.65419996]]

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

m = nn.SoftShrink()
input_ = np.array([[0.5297, 0.7871, 1.1754], [0.7836, 0.6218, -1.1542]], dtype=np.float32)
input_t = Tensor(input_, mindspore.float32)
output = m(input_t)
print(output)
# [[ 0.02969998  0.28710002  0.6754    ]
#  [ 0.28359997  0.12180001 -0.65419996]]