Ghost experiment remains if initExp fails
I believe this is an unintended consequence of https://github.com/nextml/NEXT/pull/183
Basically if initExp fails, the experiment doc still exists because it was created before initExp was called, so the experiment will still show up on the dashboard... but if you click on the link you get a Internal Server Error and error messages like:
nextbackenddocker_1 | File "/next_backend/next/api/resource_manager.py", line 222, in get_algs_for_exp_uid
nextbackenddocker_1 | for alg in args['alg_list']:
nextbackenddocker_1 | TypeError: 'NoneType' object has no attribute '__getitem__'
Probably the easiest way to fix this is to have a try/except clause around myApp.initExp that cleans up after itself if initExp fails. Any other ideas? Could this ever backfire?
Just noticed that App.initExp already has a try/except clause so this is probably what we should do:
https://github.com/nextml/NEXT/blob/master/next/apps/App.py#L96-L117