periodic eval after eval_every iteration

[AntonioCarta]

This PR enables the periodic evaluation after every eval_every iterations by setting peval_mode='iteration' in the BaseStrategy constructor. Also:

• can be easily adapted to any counter in the future, if we find it useful
• allow plugins to call eval during training
• fix bug during the restore of train/eval modes

Closes #838.

[vlomonaco]

wow nice @AntonioCarta! :)

[ContinualAI-bot]

Oh no! It seems there are some PEP8 errors! 😕
Don't worry, you can fix them! 💪
Here's a report about the errors and where you can find them:

tests/test_metrics.py:439:81: E501 line too long (81 > 80 characters)
tests/test_metrics.py:442:81: E501 line too long (81 > 80 characters)
2       E501 line too long (81 > 80 characters)

[AntonioCarta]

@AndreaCossu can you take a look? I have no idea what's going on with the metrics tests. I updated the file but I'm still getting the error. On my machine the tests are passing.

[AndreaCossu]

Can you pass me the 3 pickle files you have on your machine?

[AndreaCossu]

I cloned your repo and ran tests on the peval_iterations branch. It's strange because tests are failing in my case also on my machine (FAST_TEST=true USE_GPU=true). If your local tests pass maybe you are using different files somehow? The ones you gave me are the same I found on the branch.