automl / ParameterImportance

Parameter Importance Analysis Tool
http://www.ml4aad.org/
BSD 3-Clause "New" or "Revised" License
76 stars 19 forks source link

Example from Manual Fails #115

Closed rlecover007 closed 4 years ago

rlecover007 commented 4 years ago

Cloned the repository, navigated to examples/spear_qcp, ran pimp --scenario_file scenario.txt --history './*/runhistory.json' --modus forward-selection

which resulted in: /home/rachel/.local/lib/python3.7/site-packages/pyparsing.py:2927: FutureWarning: Possible set intersection at position 3 self.re = re.compile( self.reString ) Traceback (most recent call last): File "/home/rachel/.local/bin/pimp", line 5, in <module> from pimp.pimp import cmd_line_call File "/home/rachel/.local/lib/python3.7/site-packages/pimp/pimp.py", line 17, in <module> from smac.optimizer.objective import average_cost ModuleNotFoundError: No module named 'smac.optimizer.objective'

mlindauer commented 4 years ago

Thanks for reporting that issue. I think that happens because of the recent SMAC release where we restructured the code quite a bit. Our main developer pf pimp (@AndreBiedenkapp) is on vacation this week. So, you have to wait a bit for a fix or you downgrade your SMAC version (e.g., by using a different conda env) for running pimp.

Best, Marius

PS: I would recommend to run CAVE (https://github.com/automl/CAVE) instead of pimp because it has the nicer interface. However, since CAVE imports pimp, it will have the same issue.

rlecover007 commented 4 years ago

Thanks for pointing me towards CAVE. I'll give it a shot once I get a separate conda environment set up or whenever this issue gets fixed.

rlecover007 commented 4 years ago

What's the latest version of SMAC that's known to work with CAVE/PyImp?

mlindauer commented 4 years ago

I haven't tried it, but I would expect that SMAC version 0.11. should work with Cave/pimp.

rlecover007 commented 4 years ago

Thanks. That appears to have done the trick.

shukon commented 4 years ago

PyImp > 1.1.0 now supports SMAC 0.12.0. Since there are breaking changes introduced, please use PyImp 1.0.X for SMAC < 0.12.0.