itsmontoya
commented
1 week ago Totally @nmoniz! There is definitely a plan on the roadmap to implement generated models. I was trying to avoid coming up with a solution that would inconvenience those who want to generate on the fly. The current idea I'm thinking of is offering an Output location field to the configuration. If this is set, we can utilize generated models. Though I'm open to other ideas and suggestions as well.
As I was playing with this I realised that there isn't a way to effectively save a model after training which means we have to retrain every single time we launch the app that needs such model. While this is not very significant for small training datasets it can quickly become an issue as it grows.