Question Question regarding multiple unity instance training

Hi all, I have one question regarding multiple unity instance training. I am using unity-mlagent 1.0.8 and ml-agent python 0.16.1. So basically I want to do some hyper-parameter testing and see the optimal ones. I want to run multiple unity instances with different training configuration files, so I decide to open several unity editors with different hyper-parameter settings. I checked the docs:https://github.com/Unity-Technologies/ml-agents/blob/release_2_verified_docs/docs/Training-ML-Agents.md#training-using-concurrent-unity-instances
Even though I changed `--base-port` on command line, the ml-agent still looks for port 5004. How do you guys do multiple unity instance training? Thanks!