Differences with torch.nn.functional.log_softmax

torch.nn.functional.log_softmax

torch.nn.functional.log_softmax(
    input,
    dim=None,
    dtype=None
)

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

mindspore.ops.log_softmax

class mindspore.ops.log_softmax(
    logits,
    axis=-1,
)

For more information, see mindspore.ops.log_softmax.

Differences

PyTorch: Support the use of dim parameters and input input to implement functions that take the logits of the softmax result.

MindSpore: Support the use of axis parameters and logits input to implement functions that take the logits of the softmax result.

Categories	Subcategories	PyTorch	MindSpore	Difference
Parameters	Parameter 1	input	logits	Same function, different parameter names
	Parameter 2	dim	axis	Same function, different parameter names
	Parameter 3	dtype	-	PyTorch is used to specify the data type of the output Tensor, which is not available in MindSpore.

Code Example

import mindspore as ms
import mindspore.ops as ops
import torch
import numpy as np

# In MindSpore, we can define an instance of this class first, and the default value of the parameter axis is -1.
x = ms.Tensor(np.array([1, 2, 3, 4, 5]), ms.float32)
output1 = ops.log_softmax(x)
print(output1)
# Out:
# [-4.451912   -3.4519122  -2.4519122  -1.451912   -0.45191208]
x = ms.Tensor(np.array([[1, 2, 3, 4, 5], [5, 4, 3, 2, 1]]), ms.float32)
output2 = ops.log_softmax(x, axis=0)
print(output2)
# Out:
# [[-4.01815    -2.126928   -0.6931472  -0.12692805 -0.01814996]
#  [-0.01814996 -0.12692805 -0.6931472  -2.126928   -4.01815   ]]

# In torch, the input and dim should be input at the same time to implement the function.
input = torch.tensor(np.array([1.0, 2.0, 3.0, 4.0, 5.0]))
output3 = torch.nn.functional.log_softmax(input, dim=0)
print(output3)
# Out:
# tensor([-4.4519, -3.4519, -2.4519, -1.4519, -0.4519], dtype=torch.float64)