search

mljar / mljar-supervised

Python package for AutoML on Tabular Data with Feature Engineering, Hyper-Parameters Tuning, Explanations and Automatic Documentation
https://mljar.com
MIT License
2.99k stars 400 forks source link

Cannot load saved models #480

Open zoejyl opened 2 years ago

zoejyl commented 2 years ago

I'm sorry if this question is trivial. I tried to load one of the saved models (e.g. learner_fold_0.catboost ) in order to get some insights into the model. I tried to import as json but it didn't work. I'm wondering which encoding the models used.

pplonski commented 2 years ago

Hi, Please try to load this model with the catboost package. The code used in AutoML: https://github.com/mljar/mljar-supervised/blob/7bd8324e904a3dfa5d60bbebfbb1460d2672102c/supervised/algorithms/catboost.py#L301-L312

pplonski commented 2 years ago

I think it will be nice to add it to docs.

dvirginz commented 2 years ago

I agree. There is no documentation on saving and loading the models. As it is an AutoML framework, I suggest a generic "one-solution fit all models" schema. I.e., loading models shouldn't be done manually; instead, calling the load model on the folder of the training should load the best model for serving.

Thank you!

williamty commented 11 months ago

As it is an AutoML framework, loading models shouldn't be done manually.