Model configuration files
In addition to the information saved by tendorflow's save, it would be helpful to also save data-related settings used in that model. Examples: random number seed, segment duration, spectrogram settings such as window length, and overlap, any processing step like cropping the spectrogram at a minimum and maximum frequency, applying blur filters, etc.
This could be in a json file that would be saved alongside the files generated by tensorflow. we could add functions to package all these files into .tar.
The .tar file would be used to share a model (in the software portal, for example). A user with this file and the data would be able to recreate a model or continue training it with additional data.
To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information