Function Differences with torch.unique

torch.unique

torch.unique(
    *args,
    **kwargs
)

For more information, see torch.unique.

mindspore.ops.unique

mindspore.ops.unique(input)

For more information, see mindspore.ops.unique.

Differences

PyTorch: Deduplicate the elements in the Tensor. The parameter sorted can be set to determine whether the output is sorted in ascending order. Set the parameter return_inverse to determine whether to output the index of each element of the input Tensor in the output Tensor. Set the parameter return_counts to determine whether to output the number of each unique value in the input Tensor; set the parameter dim to specify the dimension of the unique. MindSpore does not support these functions.

MindSpore: Deduplicate the elements in the Tensor, as well as return the position index of each element of the input Tensor in the output Tensor.

Categories	Subcategories	PyTorch	MindSpore	Difference
Parameters	Parameter1	input	input	Consistent
	Parameter2	sorted	-	When sorted is True, the output Tensor is sorted in ascending order; when sorted is False, it is sorted in the original order
	Parameter3	return_inverse	-	When return_inverse is True, the index position of each element of the input Tensor in the output Tensor is returned
	Parameter4	return_counts	-	When return_counts is True, the number of each element of the output Tensor in the input Tensor is returned
	Parameter5	dim	-	Specify the dimension of unique

Code Example

# In MindSpore
import mindspore

x = mindspore.Tensor([1, 3, 2, 3], mindspore.float32)
output, idx = mindspore.ops.unique(x)
print(output)
# [1. 3. 2.]
print(idx)
# [0 1 2 1]

# In PyTorch
import torch

output, inverse_indices, counts = torch.unique(torch.tensor([1, 3, 2, 3], dtype=torch.long), sorted=True, return_inverse=True, return_counts=True)
print(output)
# tensor([1, 2, 3])
print(inverse_indices)
# tensor([0, 2, 1, 2])
print(counts)
# tensor([1, 1, 2])

# Example of using unique with dim
output, inverse_indices = torch.unique(torch.tensor([[3, 1], [1, 2]], dtype=torch.long), sorted=True, return_inverse=True, dim=0)
print(output)
# tensor([[1, 2],
#         [3, 1]])
print(inverse_indices)
# tensor([1, 0])