Improve signal processing specifications
When saving a ketos model there's an option to include an audio specification. This describes some of the signal processing applied to the raw audio, like the parameters for spectrograms computation, the sampling rate to indicate if resampling is necessary and if there's frequency cutoff.
This description is not complete, however. There are signal processing operations that can be applied and currently are not included in the specification.
A few examples are:
On top of that, it would be helpful to add information about the expected input, such as array shape ( e.g.: 1,129,94,1) and data type (e.g.: float 16, float 32, etc).
To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information