Multi-Output

Multi-Output#

split#

def split(x: 'Tensor', num_splits: 'int', axis: 'int' = 0) -> 'list':

Split a tensor into num_splits equal chunks along axis.

Parameters

  • x – Input tensor. The size along axis must be divisible by num_splits.

  • num_splits – Number of equal parts to split into.

  • axis – Axis along which to split. Default: 0.

Returns

List of num_splits tensors each with size x.shape[axis] // num_splits along axis and the same size as x in all other dimensions.

chunk#

def chunk(x: 'Tensor', chunks: 'int', axis: 'int' = 0) -> 'list':

Split a tensor into chunks chunks along axis.

The last chunk may be smaller if the axis size is not divisible by chunks.

Parameters

  • x – Input tensor.

  • chunks – Number of chunks to split into.

  • axis – Axis along which to split. Default: 0.

Returns

List of tensors. All chunks except possibly the last have size ceil(x.shape[axis] / chunks) along axis.

unbind#

def unbind(x: 'Tensor', axis: 'int' = 0) -> 'list':

Remove axis and return a list of slices along that dimension.

Analogous to Python’s list(x) applied along axis.

Parameters

  • x – Input tensor.

  • axis – Axis to remove. Default: 0.

Returns

List of x.shape[axis] tensors, each with rank one less than x.

minmax#

def minmax(x: 'Tensor') -> 'dict[str, Tensor]':

Compute both the global minimum and maximum of x.

Parameters

  • x – Input tensor of any shape.

Returns

```{‘min’` – scalar_min, ‘max’: scalar_max}`` — both are scalar tensors.

Contents