Differences with torch.amax

The following mapping relationships can be found in this file.

PyTorch APIs	MindSpore APIs
torch.amax	mindspore.ops.amax
torch.Tensor.amax	mindspore.Tensor.amax

torch.amax

torch.amax(input, dim, keepdim=False, *, out=None) -> Tensor

For more information, see torch.amax.

mindspore.ops.amax

mindspore.ops.amax(x, axis=(), keepdims=False) -> Tensor

For more information, see mindspore.ops.amax.

Differences

PyTorch: Find the maximum element of input according to the specified dim. keepdim controls whether the output and the input have the same dimension. out can get the output.

MindSpore: Find the maximum element of x according to the specified axis. The keepdims function is identical to PyTorch. MindSpore does not have out parameter. MindSpore axis has a default value, and finds the maximum value of all elements of x if axis is the default value.

Categories	Subcategories	PyTorch	MindSpore	Differences
Parameters	Parameter 1	input	x	Same function, different parameter names
	Parameter 2	dim	axis	MindSpore axis has a default value, while PyTorch dim has no default value
	Parameter 3	keepdim	keepdims	Same function, different parameter names
	Parameter 4	out	-	PyTorch out can get the output. MindSpore does not have this parameter

Code Example

# PyTorch
import torch

input = torch.tensor([[1, 2, 3], [3, 2, 1]], dtype=torch.float32)
print(torch.amax(input, dim=0, keepdim=True))
# tensor([[3., 2., 3.]])

# MindSpore
import mindspore

x = mindspore.Tensor([[1, 2, 3], [3, 2, 1]], dtype=mindspore.float32)
print(mindspore.ops.amax(x, axis=0, keepdims=True))
# [[3. 2. 3.]]