The training framework needs to be improved,
- with more flexibility in multi objective losses,
- how the training dataset is introduced over the epochs and
- finally improvements to the hyperparameter tuner that is built in.
Code has been developed organically so far, it could benefit from
- refactoring and design thinking
- configurability
Tasks TBD
The training framework needs to be improved,
Code has been developed organically so far, it could benefit from
Tasks TBD