I want env parameter for fit method of estimators

[xnaiman]

System Information

• Framework (e.g. TensorFlow) / Algorithm (e.g. KMeans): Scikit-Learn
• Framework Version: 0.20.0 (official sagemaker-scikit-learn-container)
• Python Version: 3.6
• CPU or GPU: CPU
• Python SDK Version: 1.26.0
• Are you using a custom image: No

Describe the problem

I'd like to send environment-variable(http_proxy) to docker container at the time of training, but there is no way. On the other hand, the env parameter is prepared for the transform method.

Minimal repro / logs

• Exact command to reproduce:

env = {'HTTP_PROXY': os.environ['HTTP_PROXY']}
 
sklearn = SKLearn(
    entry_point='scikit_learn_iris.py',
    train_instance_type="ml.c4.xlarge",
    role=role,
    sagemaker_session=sagemaker_session,
    hyperparameters={'max_leaf_nodes': 30}
    )

sklearn.fit({'train': train_input}, env=env)      # I want to input env paramater here

[mvsusp]

Hello @xnaiman,

SageMaker Training Service does not allow you to specify environment variables at this moment. You need to set this variable in your training script. Another alternative is to generate your own Dockerfile and set the variable there.

Thanks for using SageMaker!

[xnaiman]

@mvsusp
In my company, I need to use a proxy that requires user authentication. Therefore, I would like to avoid directly defining environment variables.
At first I will use the method you have taught, and try to make it correspond to hyperparameters argument.

[mvsusp]

Using the hyperparameters is a good alternative, @xnaiman.