Setting up eval pipeline

[bethanyconnolly]

Made changes to eval main() to allow wandb tracking of evaluation metrics

[maw501]

Do you think it might be easier to not modify main.py but instead create a new script which is the same as your current one? I'm just thinking that if we try to sync the fork we're going to have to resolve all the conflicts each time. So:

• main.py: the original file
• something.py: the new one you want to create (i.e. the same as main.py currently on this PR).

You'll also need to update the other PR here to reflect this.