Set start_iteration to zero in controller.py

[tom-dudley]

I couldn't get things to run after #23 so this might be required.

[CLAassistant]

All committers have signed the CLA.

[codelion]

Does it still work if we load from a checkpoint?

[hafeezhmha]

Oh yes! the start_iteration was being used before it's defined, it should be initialized to zero

[hafeezhmha]

Does it still work if we load from a checkpoint?

yeah, the fix works fine with checkpoints! It sets start_iteration to 0 initially, then take the actual iteration from the database before the loop starts, so it'll pick up right where it left off

[tom-dudley]

Does it still work if we load from a checkpoint?

yeah, the fix works fine with checkpoints! It sets start_iteration to 0 initially, then take the actual iteration from the database before the loop starts, so it'll pick up right where it left off

Yep that's my understanding too.

It does make me think that the multiple calls to self.database.add(initial_program) after each checkpoint restart might 'pollute' the database though. I don't believe there's a check for if code content is equal and each initial_program gets a new GUID each time thus I think it'd be overrepresented (unsure what impact that has on the underlying algorithms though).

[tom-dudley]

I think we can close this now that #25 is merged.

[codelion]

It does make me think that the multiple calls to self.database.add(initial_program) after each checkpoint restart might 'pollute' the database though. I don't believe there's a check for if code content is equal and each initial_program gets a new GUID each time thus I think it'd be overrepresented (unsure what impact that has on the underlying algorithms though).

I fixed this in #27