https://hgpu.org/?p=24776
De-specializing an HLS library for Deep Neural Networks: improvements upon hls4ml