Hi folks, As the title mentions, is it possible, in the case of a pipeline that uses cross-language transforms, to run transforms whose SDK is native to the runner directly on the runner, and only those not native to the runner in SDK harnesses?
For example, I'm currently experimenting using python xlang transforms in a Java pipeline and attempting to run it on Flink using PortableRunner. I'm expecting the java transforms to be run on flink taskmanagers directly, and only the python transforms to be delegated to the sdk harness attached to the Flink taskmanagers. However, the reality is that all transforms in the pipeline, be it Java or python, are executed in their respective SDK harnesses. Any help or info is appreciated! Regards, Lennon