🧶 Modular VAE disentanglement framework for python built with PyTorch Lightning ▸ Including metrics and datasets ▸ With strongly supervised, weakly supervised and unsupervised methods ▸ Easily configured and run with Hydra config ▸ Inspired by disentanglement_lib
Describe the bug
Schedules may unintentionally be affected by the use of validation and test data enabled in release v0.3.4 5695747c1e94420c024f1505d9b8a4b3c81ad610
These steps might increment trainer.global_step
Conversely validation and test steps will use the current value of the schedule.
This should probably instead be the original value in the config
This issue is in response to the fixes introduced in #22
Expected behaviour
I am not sure what the expected behaviour should be in some cases. Further discussion may be required.
The framework may need to check if a schedule wants its value to be used or a default value for the validation and test steps, this functionality should possibly be configurable.
Describe the bug Schedules may unintentionally be affected by the use of validation and test data enabled in release v0.3.4 5695747c1e94420c024f1505d9b8a4b3c81ad610
Conversely validation and test steps will use the current value of the schedule.
This issue is in response to the fixes introduced in #22
Expected behaviour I am not sure what the expected behaviour should be in some cases. Further discussion may be required.