Function Differences with torch.unique

torch.unique

torch.unique(
    input,
    sorted=True,
    return_inverse=False,
    return_counts=False,
    dim=None
)

For more information, see torch.unique.

mindspore.ops.Unique

class mindspore.ops.Unique(*args, **kwargs)(x)

For more information, see mindspore.ops.Unique.

Differences

PyTorch: By setting relevant parameters, determines whether to sort the output, to return indices of elements in the input corresponding to the output tensor, to return counts for each unique element.

MindSpore: Outputs all unique elements in ascending order, and returns indices of elements in the input corresponding to the output tensor.

Code Example

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

# In MindSpore, the tensor containing unique elements in ascending order.
# As well as another tensor containing the corresponding indices will be directly returned.
x = ms.Tensor(np.array([1, 2, 5, 2]), ms.int32)
unique = ops.Unique()
output, indices = unique(x)
print(output)
print(indices)
# Out：
# [1 2 5]
# [0 1 2 1]

# In torch, parameters can be set to determine whether to output tensor containing unique elements in ascending order.
# As well as whether to output tensor containing corresponding indices.
x = torch.tensor([1, 2, 5, 2])
output, indices = torch.unique(x, sorted=True, return_inverse=True)
print(output)
print(indices)
# Out：
# tensor([1, 2, 5])
# tensor([0, 1, 2, 1])