unique¶

dragon.unique(
  inputs,
  return_inverse=False,
  return_counts=False,
  **kwargs
)[source]¶

Return the unique elements of input.

If return_inverse, return the extra index where input mapping to:

x = dragon.constant([1, 2, 3, 2])
y, index = dragon.unique(x, return_inverse=True)
print(y)  # [1, 2, 3]
print(index)  # [0, 1, 2, 1]

If return_counts, return the extra counts of output:

x = dragon.constant([1, 2, 3, 2])
y, counts = dragon.unique(x, return_counts=True)
print(y)  # [1, 2, 3]
print(counts)  # [1, 2, 1]

Parameters:

• inputs (dragon.Tensor) – The input tensor.
• return_inverse (bool, optional, default=False) – Return the inverse index or not.
• return_counts (bool, optional, default=False) – Return the counts or not.

Returns:

• dragon.Tensor – The output tensor.
• dragon.Tensor, optional – The inverse index tensor.
• dragon.Tensor, optional – The counting tensor.