Fine Tuning

April 8, 2020 ยท View on GitHub

Checkout the script train.py. You will notice several blocks that check whether we want to perform fine tuning. If that is the case, the script will look for a registered model in the AML Model Registry, instead of training from scratch.

Note the additional input argument freeze_layers to the script. This allows us to explore which layers get to learn during fine-tuning.