Typically, you want to optimize the use of a large VM hosting
your notebook session by parallelizing the different workloads that
are part of the machine learning (ML) lifecycle. For example,
doing extract-transform-load (ETL) operations, data preparation,
feature engineering, or model training can be parallelized. You
don't want to use a
VM.Standard2.24 shape with 24 ocpus and realize that your code
is only using one of those CPUs. That would be a costly and
suboptimal use....