The filter params are stored and easily accessible via a wandb api, which can take care of all our bookkeeping for various training runs. So it would be nice to have a script to take a wandb run, and for a chosen epoch plot visualisations of the filters. We could also experiment with storing these visualisations at certain intervals automatically using wandb, as long as it doesn't get too bulky.
The filter params are stored and easily accessible via a wandb api, which can take care of all our bookkeeping for various training runs. So it would be nice to have a script to take a wandb run, and for a chosen epoch plot visualisations of the filters. We could also experiment with storing these visualisations at certain intervals automatically using wandb, as long as it doesn't get too bulky.