[FEA] Reshape Operator
tylera-nvidia opened this issue · comments
Tyler Allen commented
The MatX API currently lacks a way to logically combine tensors into a tensor of different size/rank in the manner of stack . A Reshape operator would provide flexibility to build further operations on a new combined set of data before it is aggregated and written to memory. This also keeps the allocation, sizing, and management of the final tensor in the hands of the user, without the need of intermediate storage/operations.