take

paddle. take ( x: Tensor, index: Tensor, mode: Literal['raise', 'wrap', 'clip'] = 'raise', name: str | None = None ) → Tensor [source]

Returns a new tensor with the elements of input tensor x at the given index. The input tensor is treated as if it were viewed as a 1-D tensor. The result takes the same shape as the index.

Parameters

• x (Tensor) – An N-D Tensor, its data type should be int32, int64, float32, float64.

• index (Tensor) – An N-D Tensor, its data type should be int32, int64.

• mode (str, optional) –

  Specifies how out-of-bounds index will behave. the candidates are 'raise', 'wrap' and 'clip'.

  • 'raise': raise an error (default);

  • 'wrap': wrap around;

  • 'clip': clip to the range. 'clip' mode means that all indices that are too large are replaced by the index that addresses the last element. Note that this disables indexing with negative numbers.

• name (str|None, optional) – Name for the operation (optional, default is None). For more information, please refer to api_guide_Name.

Returns

Tensor, Tensor with the same shape as index, the data type is the same with input.

Examples

>>> import paddle >>> x_int = paddle.arange(0, 12).reshape([3, 4]) >>> x_float = x_int.astype(paddle.float64) >>> idx_pos = paddle.arange(4, 10).reshape([2, 3]) # positive index >>> idx_neg = paddle.arange(-2, 4).reshape([2, 3]) # negative index >>> idx_err = paddle.arange(-2, 13).reshape([3, 5]) # index out of range >>> paddle.take(x_int, idx_pos) Tensor(shape=[2, 3], dtype=int64, place=Place(cpu), stop_gradient=True, [[4, 5, 6],  [7, 8, 9]]) >>> paddle.take(x_int, idx_neg) Tensor(shape=[2, 3], dtype=int64, place=Place(cpu), stop_gradient=True, [[10, 11, 0 ],  [1 , 2 , 3 ]]) >>> paddle.take(x_float, idx_pos) Tensor(shape=[2, 3], dtype=float64, place=Place(cpu), stop_gradient=True, [[4., 5., 6.],  [7., 8., 9.]]) >>> x_int.take(idx_pos) Tensor(shape=[2, 3], dtype=int64, place=Place(cpu), stop_gradient=True, [[4, 5, 6],  [7, 8, 9]]) >>> paddle.take(x_int, idx_err, mode='wrap') Tensor(shape=[3, 5], dtype=int64, place=Place(cpu), stop_gradient=True, [[10, 11, 0 , 1 , 2 ],  [3 , 4 , 5 , 6 , 7 ],  [8 , 9 , 10, 11, 0 ]]) >>> paddle.take(x_int, idx_err, mode='clip') Tensor(shape=[3, 5], dtype=int64, place=Place(cpu), stop_gradient=True, [[0 , 0 , 0 , 1 , 2 ],  [3 , 4 , 5 , 6 , 7 ],  [8 , 9 , 10, 11, 11]])