pad¶

dragon.pad(
  inputs,
  pads,
  mode='constant',
  value=0,
  **kwargs
)[source]¶

Pad the input according to the given sizes.

pads should be a sequence with \(N\) tuples, where \(N\) is the number of input dimensions:

x = dragon.ones(shape=(2, 3))
print(dragon.pad(x, [[0, 1], [1, 0]]))  # Ok, (2, 3) -> (3, 4)
print(dragon.pad(x, [[0, 1]]))  # Wrong

Following padding modes are supported:

x = dragon.constant([[1, 2, 3], [4, 5, 6]])
# ConstantPad
print(dragon.pad(x, [[0, 1], [1, 0]], 'constant', 9))
# ReflectPad
print(dragon.pad(x, [[0, 1], [1, 0]], 'reflect'))
# EdgePad
print(dragon.pad(x, [[0, 1], [1, 0]], 'edge'))

Parameters:

• inputs (dragon.Tensor) – The input tensor.
• pads (Sequence[Tuple[int], Tuple[dragon.Tensor]]) – The padding begins and ends.
• mode ({'constant', 'reflect', 'edge'}, optional) – The padding mode.
• value (number, optional, default=0) – The value used in 'constant' mode.

Returns:

dragon.Tensor – The output tensor.