I use Knime to build deep learning models and and very happy with the functionality and ease of use. I run Knime on my MacBook Pro with 32 GB RAM. While the CPU performance is acceptable for many of the deep learning models that I have built, I would like to create some CNN models where support for the Radeon Pro 560X 4 GB on my Mac should help with performance. PlaidML has support for OpenCL and Apple Metal. I am wondering if you have any plans to integrate PlaidML as a backend in your Keras integration? This would allow Mav users to take advantage of their additional GPU.