Models are designed for a specific set of feature extraction parameters. But, so far the parameters are only stored in the parameters.json file. So, we should save the parameters used for the design and training of the model.
This could be implemented by several ways, three options are:
Save a json file in the model folder. This is similar to how we save the parameters in the features folder.
Save the FeatureExtractor instance (pickle file) in the model folder. This method is perhaps more opaque to users, but it seems like a good solution.
A third option could be to save the ModelContainer instance in the model folder. To do this, we should include the FeatureExtractor instance in the ModelContainer. This change library usage, so we must discuss more about this.
Models are designed for a specific set of feature extraction parameters. But, so far the parameters are only stored in the parameters.json file. So, we should save the parameters used for the design and training of the model.
This could be implemented by several ways, three options are:
Save a json file in the model folder. This is similar to how we save the parameters in the features folder.
Save the FeatureExtractor instance (pickle file) in the model folder. This method is perhaps more opaque to users, but it seems like a good solution.
A third option could be to save the ModelContainer instance in the model folder. To do this, we should include the FeatureExtractor instance in the ModelContainer. This change library usage, so we must discuss more about this.