Reputation: 2607
I am trying to go through the following tutorial published here but get the error below when I run these lines fo code:
run = exp.submit(est)
run.wait_for_completion(show_output=True)
ERROR:
"message": "Could not import package \"azureml-dataprep\". Please ensure it is installed by running: pip install \"azureml-dataprep[fuse,pandas]\""
However, I have already installed the required packages:
I am running this through Jupyter Notebooks in an Anacoda Python 3.7 environment.
UPDATE
Tried creating a new conda environment as specified here but still get the same error.
conda create -n aml python=3.7.3
After installing all the required packages, I am able to reproduce the exeception by executing the following:
Upvotes: 1
Views: 5279
Reputation: 501
Sorry for this. Take a look at the Jupyter Notebook version of the same tutorial: https://github.com/Azure/MachineLearningNotebooks/blob/master/how-to-use-azureml/ml-frameworks/tensorflow/deployment/train-hyperparameter-tune-deploy-with-tensorflow/train-hyperparameter-tune-deploy-with-tensorflow.ipynb
When configuring estimator, you need to specify the pip packages u wanna install on the remote compute. In this case, azureml-dataprep[fuse, blob]. Installing the package to your local computer is not useful since the training script is executed on the remote compute target, which doesn't have the required package installed yet.
est = TensorFlow(source_directory=script_folder,
script_params=script_params,
compute_target=compute_target,
entry_script='tf_mnist.py',
use_gpu=True,
pip_packages=['azureml-dataprep[pandas,fuse]'])
Can you pls try the fix and let us know whether it solves your issue :) In the mean time, I will update the public documentation to include pip_packages in estimator config.
Upvotes: 1
Reputation: 222722
Have you gone through the known issues and Troubleshooting
page?. It is mentioned as one of the known issue.
Error message: ERROR: No matching distribution found for azureml-dataprep-native
Anaconda's Python 3.7.4 distribution has a bug that breaks azureml-sdk install. This issue is discussed in this GitHub Issue This can be worked around by creating a new Conda Environment using this command:
Upvotes: 1