Using Machine Learning to Reconstruct Seismic Motion

I am working on a large-scale problem for the first time and would appreciate any form of assistance in understanding concepts that would efficiently allow me to maximize the use of an amazing resource such as the supercomputers at PSC Bridges. What I feel bad about is accidentally using a lot of resources that might only require half of the allocated resource, leading to a waste of allocated computing time and resources, all because I do not understand the allocation process properly. I primarily use Jupyter Labs and TensorFlow for machine learning applications. Additionally, I am responsible in my lab for supervised machine learning and teaching other people, which is why I would appreciate this support.