This is cool
Next step is tuning your post training pipeline to optimize your model for your chosen FPGA architecture...