Versioned name: DynamicTranspose-1
Short description: DynamicTranspose operation reorders the input tensor dimensions. In DynamicTranspose, the target shape order is given as an input tensor at runtime. It’s useful when the target order is unknown during the operator creation. Use DynamicTranspose if order is not constant or is not available until runtime. Otherwise, use StaticTranspose.
Detailed description: DynamicTranspose operation reorders the input tensor dimensions. Source indices and destination indices are bound by the formula:
The input shape is [input.shape(0), input.shape(1), ……, input.shape(N-1)], the output shape is [input.shape(order), input.shape(order), …, input.shape(order[N-1])]. Output tensor may have a different memory layout with input tensor. DynamicTranspose is not guaranteed to return a view or a copy when input tensor and output tensor can be inplaced, users should not depend on this behavior.
data- the tensor to be transposed. Required.
order- the permutation applied to the axes of the input shape. It must be a vector of elements with T2 type and shape [N], where N is the rank of
data. If an empty list  is specified, then axes will be inverted to ([N-1,…,1,0]). A negative number means counting from last to the first axis. Required.
1: A tensor transposed from input
T: f32, f16, bf16.
Note: Inputs and outputs have the same data type denoted by T. For example, if input is f32 tensor, then all other tensors have f32 data type.