How to add python executable path for sagemaker SKLearn python SDK? - scikit-learn

Question

When I run train.py via SKLearn estimater in sagemaker python SDK it runs using default python executable path of sagemaker which is /miniconda3/bin/python
I want to run using another python executable path which is /miniconda3/envs/interplay-env/bin/python3
How to set this in sagemaker SKLearn estimator?
Below is current configuration of sagemaker estimator
sklearn_estimator = SKLearn(
entry_point=train.py,
role=role,
instance_count=1,
instance_type="ml.c5.xlarge",
framework_version=FRAMEWORK_VERSION,
base_job_name=training_job_name,
hyperparameters={
"parameters_dict": parameters_dict1,
"features": features,
"target": target,
"project_path":project_path,
"bucket_name":bucket_name,
}
)
I tried setting the executable path in entry_point like below. But it doesn't work.
sklearn_estimator = SKLearn(
entry_point=["/miniconda3/envs/interplay-env/bin/python3",train_file_name],
role=role,
instance_count=1,
instance_type="ml.c5.xlarge",
framework_version=FRAMEWORK_VERSION,
base_job_name=training_job_name,
hyperparameters={
"parameters_dict": parameters_dict1,
"features": features,
"target": target,
"project_path":project_path,
"bucket_name":bucket_name,
}
)

Answer 1

Not sure why you would want to do that. By Default SageMaker runs the code in a container and provides a separate conda environment for your code to run. If you need additional libraries you can have a requirements.txt and add the dependent libraries into it. Later you will add another attribute in your SKlearn Estimator called source_dir point to a folder that will contain your entrypoint script and the requirements.txt