Train on load?

[mzed]

When an object is instantiated with an existing dictionary, should it automatically train? What order do these things happen when a patch is loaded?

[francesco-di-maggio]

This is an interesting feature. It would be nice indeed to have rapid train automatically a model when instantiated with an existent dictionary file name. Alternatively, in Max one could dynamically read a dict file (or use a "loadmess #1" > prepend read), inform the rapid object prepending by the "trainingDict" message, and then train rapid. Like so. Food for thoughts.

[mzed]

An object with an @trainingDict attribute will train on that with just a train message. So, "train" could be triggered by loadbang/loadmess.

I'm not 100% sure of the ordering (when is that dictionary ready?), so there would need to be some testing. Also, a really huge dictionary could cause a substantial training time. These are my concerns. If read/write models is working properly, then maybe that's a better solution.

[francesco-di-maggio]

Good point. Maybe the and objects could come in hand in this case? So far I used this workflow of dict reading > rapid addressing > train message deferring, and never had issues. But certainly something to test further. On another prospective, objects like buffer~ seem to implement what - I think - reflects the intentions behind this issue: i.e. automatically load target content when instantiated with an existent buffer name...?