swapLast2Axes

Hardware Model	Supported or Not
Atlas 200I/500 A2 inference products	Not supported
Atlas inference products	Not supported
Atlas training products	Not supported
Atlas A2 training products / Atlas A2 inference products	Supported
Atlas A3 training products / Atlas A3 inference products	Supported

API function

swapLast2AxesGetWorkspaceSize: calculates the workspace size required by the swapLast2Axes operator.

To use the swapLast2Axes operator, you need to call the swapLast2AxesGetWorkspaceSize API to obtain the workspace size required for calculation and the executor that contains the operator calculation process, and then call the swapLast2Axes API to perform calculation.
swapLast2Axes: swaps the last two dimensions of a tensor.

Formula:

The calculation formula of swapLast2Axes is $\text{[math]}$ , where b indicates the batch number of data, i indicates the row number of the input data, and j indicates the column number of the input data.

Example:

Example 2:

inTensor:

[[[ 0.+0.j, 1.+0.j, 2.+0.j], 
[ 3.+0.j, 4.+0.j, 5.+0.j]],

[[ 6.+0.j, 7.+0.j, 8.+0.j], 
[ 9.+0.j, 10.+0.j, 11.+0.j]]],

After the swapLast2Axes operator is called, the outTensor is as follows:

[[[ 0.+0.j, 3.+0.j], 
[ 1.+0.j, 4.+0.j], 
[ 2.+0.j, 5.+0.j]],

[[ 6.+0.j, 9.+0.j], 
[ 7.+0.j, 10.+0.j], 
[ 8.+0.j, 11.+0.j]]].

AsdSip::AspbStatus swapLast2AxesGetWorkspaceSize(const aclTensor *inTensor, aclTensor *outTensor, size_t *size)
AsdSip::AspbStatus swapLast2Axes(const aclTensor *inTensor, aclTensor *outTensor, void *stream,void *workspace = nullptr)

swapLast2AxesGetWorkspaceSize

Parameter	Input/Output	Type	Description
inTensor	Input	const aclTensor *	In the formula, inTensor indicates the tensor on the device. The input dimension is 2 or 3. The maximum number of input elements is 3600000000. The data type must be COMPLEX64 and the data format must be ND.
outTensor	Input	aclTensor *	outTensor: tensor on the device. The data type must be COMPLEX64. If the shape of inTensor is [k, x, y], the shape of outTensor is [k, y, x], the data type must be the same as that of inTensor. The data format can be ND.
workspaceSize	Input	size_t *	Workspace required by the swapLast2Axes operator.

swapLast2Axes

Parameter	Input/Output	Type	Description
inTensor	Input	const aclTensor *	In the formula, inTensor indicates the tensor on the device. The input dimension is 2 or 3. The maximum number of input elements is 3600000000. The data type must be COMPLEX64 and the data format must be ND.
outTensor	Output	aclTensor *	outTensor: tensor on the device. The data type must be COMPLEX64. If the shape of inTensor is [k, x, y], the shape of outTensor is [k, y, x], the data type must be the same as that of inTensor. The data format can be ND.
stream	Input	void*	NPU execution flow.
workspaceSize	Input	void *	Workspace required by the swapLast2Axes operator.

For details about the return values, see Return Value.

swapLast2AxesGetWorkspaceSize: none.
swapLast2Axes
- The maximum number of input elements is 3600000000 (within [60000, 60000]).
- If the input shape of the operator is [k, x, y], the output shape of the operator is [k, y, x].
- The operator does not support the calculation for more than three dimensions.

For details about the operator calling example, see swapLast2Axes.

Parent topic: Base