search

unit8co / darts

A python library for user-friendly forecasting and anomaly detection on time series.
https://unit8co.github.io/darts/
Apache License 2.0
7.91k stars 857 forks source link

Train a model, but don't save model artifacts #510

Closed jonhilgart22 closed 2 years ago

jonhilgart22 commented 2 years ago

Is your feature request related to a current problem? Please describe. A clear and concise description of what the problem is.

I'd like to train a Nbeats and TCN model but not save the model artifacts. I'm running this code on AWS Lambda which has size restrictions for these model files.

Describe proposed solution A clear and concise description of what the library should provide to solve missing functionality.

Can we pass in a param here https://github.com/unit8co/darts/blob/484a184a97be37e64ed62f6fa6fad2bdef898adf/darts/models/forecasting/torch_forecasting_model.py#L859 indicating whether save=True or save=False ?

Describe potential alternatives A clear and concise description of any alternative solutions or existing features that might solve it.

Additional context Add any other context or screenshots about the feature request here.

dennisbader commented 2 years ago

Hi @jonhilgart22 and thanks for pointing this out.

Currently, the models are saved automatically without the option to bypass it.

We added this feature to our to-do list!

hrzn commented 2 years ago

Since v0.13.0, checkpoints are not saved by default anymore (you can still save them by passing save_checkpoints=True to the model).