-
Paul Fultz II authored
Converts can be inserted when the scales and input differ in the onnx file(we are already doing this implicit conversion in the ref implementation). This will also improve the compile-time of quantizelinear.hpp since we can remove the nested visit method.
f6e22d56