Currently the user is able to run DeepSSM augmentation with set split percentages (ie: 25%, 15%) and then change those values in the studio UI and run training, which will pull the updated values. This means there can be a discrepancy between the two tasks split percentages, causing unintended behavior and potentially errors.
There should be some form of enforcement on when the split values are saved. Simply passing the job type to store_params and adding a condition to only save if the task is Augmentation could suffice
Currently the user is able to run DeepSSM augmentation with set split percentages (ie: 25%, 15%) and then change those values in the studio UI and run training, which will pull the updated values. This means there can be a discrepancy between the two tasks split percentages, causing unintended behavior and potentially errors.
There should be some form of enforcement on when the split values are saved. Simply passing the job type to
store_params
and adding a condition to only save if the task is Augmentation could suffice