[Bug Report] RuntimeError: Dataset not found. You can use download=True to download it for pytorch minist horovod

I got `RuntimeError: Dataset not found. You can use download=True ` when running [MNIST Training using PyTorch](https://sagemaker-examples.readthedocs.io/en/latest/sagemaker-python-sdk/pytorch_horovod_mnist/pytorch_mnist_horovod.html).

I am trying to change the Python version to 3.8 as a workaround but got new error:
```
ErrorMessage "RuntimeError: CUDA error: an illegal memory access was encountered
CUDA kernel errors might be asynchronously reported at some other API call,so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1."
```

Any hints appreciate!

Start with the linked MNIST Training using PyTorch example at sagemaker-python-sdk/pytorch_horovod_mnist/pytorch_mnist_horovod.html and reproduce the reported dataset error. Compare the original run with the Python 3.8 attempt and CUDA error, then document the confirmed cause and a reproducible resolution for the example.

Written by the indexing model from the issue text.