Developer Guide for Intel® Data Analytics Acceleration Library 2019 Update 2

Reshape Forward Layer

The forward reshape layer generates a tensor from input argument X of size n 1 x n 2 x ... x n p with modified size of dimensions m 1 x m 2 x ... x m q without modifying data and its order.

Problem Statement

Given the dimension sizes of the output tensor m 1 x m 2 x ... x m q , such that



Note

There are two reserved values of m i :

  • 0 means m i = n i , valid only when ip

  • undefinedDimensionSize means m i is calculated to make the product of all mi equal to the product of all n i, valid only if used once. Corresponding m i calculated as follows:



The q-dimensional result tensor Y of size m 1 x ... x m q contains the same data from X in unmodified order.