Load the saved model for inference

[henjoe]

Hi, first of all thank you for wonderful job!

I justbwant to clarify things since I dont see this in the documentations.

If let say I trained a model, as per documentation it is automatically saved (set on default folder), now I stop my python environment and then run it again, what I want is to use the saved model so I can easily inference with it without fitting the model again.

For example

predictor = AutoML(load='path'):

Does this supported now?

Thank you.

[pplonski]

Hi @henjoe,

Yes, it is supported. Please use results_path as argument in the constructor. Example:

automl = AutoML(results_path="my_automl_models")
predictions = automl.predict(X_test)

May I ask what is your use case for AutoML? :)

[henjoe]

Hi thanks for a quick reply!
For my use case, I wanna use it on Industrial Automation (which I am on for several years).

Quick question, once I do the

predictor = AutoML(result_path)

I believed AutoML trained multiple models

1.will it be possible to define that only use the model with higher accuracy in the predictor?

2. Everytime I infer to the model, does it load from the disk (which will take some time) or on memory (faster)?

[pplonski]

AutoML will always use the best model for computing predictions. Right now there is no option to select other model. Models are only loaded once, when you construct the AutoML object. They are stored in RAM memory when computing predictions.