Change the model source path in minio during the experiment and serving job

Both Triton Inference Server and TensorFlow serving will get the model name and model version from the model URL however the model version in the experiment and management is different. It will be confused when inferencing this model.

https://docs.google.com/presentation/d/13OwvqNst35p_kV_iTHv40fdASkc4XvmLGd9cF1ZicSc/edit?usp=sharing