Description

Current:

net.eval()
with torch.no_grad():
    pass
net.train()

Proposed:

with eval_mode(net):
    pass

This is also useful when the train/eval state of a network could be one or the other, yet we want it to temporarily be in eval mode and return to original state at the end (e.g., gradcam).

To avoid conflicts, I haven't updated gradcam or occlusion sensitivity until #1381 is merged. I'll also update the tutorials.

Status

Ready

Types of changes

• Non-breaking change (fix or new feature that would not break existing functionality).
• New tests added to cover the changes.
• Integration tests passed locally by running ./runtests.sh --codeformat --coverage.
• Quick tests passed locally by running ./runtests.sh --quick.
• In-line docstrings updated.
• Documentation updated, tested make html command in the docs/ folder.