Training from warm start with multiple GPUs

[ohinds]

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)

Summary

Enable training from a warm start. The tensorflow distributed strategy is now maintained as a member variable of the estimator so a model can be loaded and trained with the same strategy object.

Resolves #239

Checklist

• I have added tests to cover my changes
• I have updated documentation (if necessary)

Acknowledgment

• I acknowledge that this contribution will be available under the Apache 2 license.