Skip to main content
Version: 10.2.8

Train and execute custom Python model

note

For details of AutoML SDK configuration, refer to AutoML SDK workflow.

Train model​

To train a custom Python model, follow the steps below:

  1. Configure paths in ModelTrainingRunner and run it. If you don't want to implement cross-validation for Python models, in the training configuration, add the testSetDir parameter.

  2. Specify any additional files for model execution in the resources parameter of LocalTrainingConfiguration as shown below:

    LocalTrainingConfiguration configuration = LocalTrainingConfiguration.builder()
        .id(modelId)
        .inputDir(trainingSet)
        .outputDir(outputDirPath)
        .testSetDir(testSetDirPath)
        .resource("models", resourcesDirPath) //add this parameter if you have some additional resources for model execution
        .parameters(parameters)
        .build();

Execute model​

Configure the paths to the input, trained model, and output directories in the ModelExecutionRunner class and run the model.

Import model to Control Tower​

For importing models to Control Tower, use the following instructions:

TROUBLESHOOTING

In case you have issues when running a Python model, read the Issues related to incorrect Python configuration guide.