rot90

paddle. rot90 ( x, k=1, axes=[0, 1], name=None ) [source]

Rotate a n-D tensor by 90 degrees. The rotation direction and times are specified by axes and the absolute value of k. Rotation direction is from axes[0] towards axes[1] if k > 0, and from axes[1] towards axes[0] for k < 0.

Parameters
  • x (Tensor) – The input Tensor(or LoDTensor). The data type of the input Tensor x should be float16, float32, float64, int32, int64, bool. float16 is only supported on gpu.

  • k (int, optional) – Direction and number of times to rotate, default value: 1.

  • axes (list|tuple, optional) – Axes to rotate, dimension must be 2. default value: [0, 1].

  • name (str, optional) – The default value is None. Normally there is no need for user to set this property. For more information, please refer to Name .

Returns

Tensor, Tensor or LoDTensor calculated by rot90 layer. The data type is same with input x.

Examples

>>> import paddle

>>> data = paddle.arange(4)
>>> data = paddle.reshape(data, (2, 2))
>>> print(data.numpy())
[[0 1]
 [2 3]]

>>> y = paddle.rot90(data, 1, [0, 1])
>>> print(y.numpy())
[[1 3]
 [0 2]]

>>> y= paddle.rot90(data, -1, [0, 1])
>>> print(y.numpy())
[[2 0]
 [3 1]]

>>> data2 = paddle.arange(8)
>>> data2 = paddle.reshape(data2, (2,2,2))
>>> print(data2.numpy())
[[[0 1]
  [2 3]]
 [[4 5]
  [6 7]]]

>>> y = paddle.rot90(data2, 1, [1, 2])
>>> print(y.numpy())
[[[1 3]
  [0 2]]
 [[5 7]
  [4 6]]]