I added gemm (aAB + bC, where a, b are scalars, and A B C are tensors) nodes to the onnx reader because that's what the pytorch exporter uses for fully connected layers.
I also changed the reader to include scaling objects and input bounds parameters.
I added gemm (
aAB + bC
, wherea, b
are scalars, andA B C
are tensors) nodes to the onnx reader because that's what the pytorch exporter uses for fully connected layers.I also changed the reader to include scaling objects and input bounds parameters.