search

prio-data / views_pipeline

VIEWS forecasting pipeline for monthly prediction runs. Includes MLops and QA for all models/ensembles.
Other
3 stars 3 forks source link

Create hydranet main #26

Closed Polichinel closed 5 months ago

Polichinel commented 5 months ago

Created a main.py for purple alien.

But more importantly, this is to be seen as a suggestion to a more general structure. I.e. both how the main is structured, the use of agr parse, and the associated common_utils. But also further the way that the hierarchic of scripts main.pt<-model_run_handlers<-model_run_managers<-(training, testing, forecasting scripts respectively) ids set up.

We should think about how much of this we can abstract out to common_utils. Of course, the deeper we go into these hierarchies the less we should expect the code to be generalizable.

Thanks in advance for any notes and feedback!

sarakallis commented 5 months ago

Looking great! 👏 Put a lot of little comments in the code, most are about how intuitive I find the naming and organization of code. And I think you should have kept a pipeline function in main. Btw I really like the small utils modules! Super neat and something to strive for.

The only other thing is that I am missing a model_config with these parameters that can be useful for future batch processing, meta tools, etc.

Also, reminder to update your utils_dataloaders to views6 and probably rename your queryset to something other than simon_tests 🤠