A low-priority feature request:
Pipedal running a Raspberry Pi 4 could make good use of slimmed models, given that in can only run two NAM instances easily. It would therefore be a nice feature if TooB Neural Amp Modeler could provide a dropdown in it's user-interface that allows selection of Full, or Light, or Pico &c models.
If the module metadata were exposed in DSP APIs, I could add a UI component to TooB Neural Amp Modeller that allows selection of slimmed models, if present in the model file.
Data I would need: a vector of doubles containing max_value metadata for each submodule in a slimmable model. (Or an enumeration of 'max_value' values in the entire DSP tree, I suppose. :-/
Failing that, I guess I could pre-load the config file and extract values out of the json model, although I'm guessing that slimmable parameters might be present elsewhere in the model than just the top-level slimmable container model.
A low-priority feature request:
Pipedal running a Raspberry Pi 4 could make good use of slimmed models, given that in can only run two NAM instances easily. It would therefore be a nice feature if TooB Neural Amp Modeler could provide a dropdown in it's user-interface that allows selection of Full, or Light, or Pico &c models.
If the module metadata were exposed in DSP APIs, I could add a UI component to TooB Neural Amp Modeller that allows selection of slimmed models, if present in the model file.
Data I would need: a vector of doubles containing
max_valuemetadata for each submodule in a slimmable model. (Or an enumeration of 'max_value' values in the entire DSP tree, I suppose. :-/Failing that, I guess I could pre-load the config file and extract values out of the json model, although I'm guessing that slimmable parameters might be present elsewhere in the model than just the top-level slimmable container model.