The new experiment should try to reproduce the execution modes of BERT (https://github.com/google-research/bert): pretraining, fine-tuning, inference. Ideally we'd like to do this in a Java runtime but may have to use python if the profiler isn't available in Java.
We want to be able to apply jcarbon to ML systems, so we want to build an experiment that uses tensorflow. Previously done with eflect here: https://github.com/pl-eflect/eflect/blob/main/experiments/src/eflect/experiments/InceptionDriver.java
The new experiment should try to reproduce the execution modes of BERT (https://github.com/google-research/bert): pretraining, fine-tuning, inference. Ideally we'd like to do this in a Java runtime but may have to use python if the profiler isn't available in Java.