PyTorch HPO example failing

The [PyTorch HPO example](https://github.com/awslabs/amazon-sagemaker-examples/tree/master/hyperparameter_tuning/pytorch_mnist) does not work and fails with error:

Failure reason:
AlgorithmError: ExecuteUserScriptError: Command "/opt/conda/bin/python mnist.py --backend gloo --batch-size 256 --epochs 6 --lr 0.04926795600136902" Traceback (most recent call last): **File "mnist.py", line 5, i to n import sagemaker_containers ModuleNotFoundError: No module named 'sagemaker_containers'**

Problem is happening for PyTorch versions > 1.3.1 possibly due to https://github.com/aws/sagemaker-pytorch-training-toolkit/issues/189

Manually adding `!pip install sagemaker-containers` to the SM notebook also does not help. Checked for PyTorch 1.6.0. 1.5.0, 1.5.1

Start in hyperparameter_tuning/pytorch_mnist and reproduce the PyTorch HPO example with the affected PyTorch versions, focusing on the mnist.py import shown in the traceback. Check the related sagemaker-pytorch-training-toolkit issue and verify that the example runs without the sagemaker_containers ModuleNotFoundError.

Written by the indexing model from the issue text.