Fixing container path for source files when using local_session

[bearpelican]

Issue #, if available:
Fixes: No such file or directory error when running estimator fit/predict with a local sagemaker session and a local script.

When running sagemaker in local mode, sagemaker_submit_directory is set to the the local environment path (Ex: file:///home/ec2-user/mnist/src'), when it should be set to the container path - '/opt/ml/code'. The container path is mounted in src/local/image.py.
The container does not have access to the local environment path, which is why a No such file or directory error happens.

Minimal example that breaks:
https://gist.github.com/bearpelican/440d10f28003c3653c7866c129418ad7

This was taken directly from the sagemaker example notebook:
https://github.com/awslabs/amazon-sagemaker-examples/tree/master/sagemaker-python-sdk/pytorch_mnist

Description of changes:

Merge Checklist

Put an x in the boxes that apply. You can also fill these out after creating the PR. If you're unsure about any of them, don't hesitate to ask. We're here to help! This is simply a reminder of what we are going to look for before merging your pull request.

• I have read the CONTRIBUTING doc
• I have added tests that prove my fix is effective or that my feature works (if appropriate)
• I have updated the changelog with a description of my changes (if appropriate)
• I have updated any necessary documentation (if appropriate)

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

[codecov-io]

Codecov Report

Merging #499 into master will increase coverage by 0.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #499      +/-   ##
==========================================
+ Coverage   92.77%   92.79%   +0.01%     
==========================================
  Files          71       71              
  Lines        5359     5371      +12     
==========================================
+ Hits         4972     4984      +12     
  Misses        387      387

Impacted Files	Coverage Δ
src/sagemaker/local/image.py	88.53% <100%> (+0.4%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4bac185...81cdfd4. Read the comment docs.

[bearpelican]

Not sure if this is the best way to fix the bug, but I wanted to put this here as an issue and proposed fix. If anyone has any other ideas, please feel free to submit a new PR and close this one!

[laurenyu]

Thanks for the contribution! Would you mind adding some tests?

[bearpelican]

Yup will do

[bearpelican]

Updated commit - 5f54934 with extended unit tests

[laurenyu]

Thanks for updating! There are a couple flake8 errors:

flake8 runtests: commands[0] | flake8
./src/sagemaker/local/image.py:111:1: W293 blank line contains whitespace
./src/sagemaker/local/image.py:338:1: W293 blank line contains whitespace

[bearpelican]

Cool. Should be fixed now!

[laurenyu]

an integ test is failing because it expects this hyperparameter to be JSON (at least for the SageMaker deep learning framework images)

[bearpelican]

Ah good catch! Updated hyperparameters with json conversion

[bearpelican]

Unable to run integration tests on my machine, so please let me know if there are any more errors on those tests

[laurenyu]

@bearpelican the integ tests look fine now, but could you add the changelog entry to a new 1.16.3 section? (1.16.2 got released yesterday)