-
Notifications
You must be signed in to change notification settings - Fork 50
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fix sampler bugs, cut down NCMC testing time, add alkanes test system #109
Conversation
I've also added an alkanes testsystem. |
Among the errors (this happens a few times and was also happening in #92): The only other things that seem to be wrong have to do with TopologyProposals taking different inputs and having different attributes now, which will be fixed in #93 I think |
I had totally missed that! Hopefully fixed now.
Is that something I can fix in this PR? |
The other failures seem to occur in
@pgrinaway: Any chance we could get in a quick fix for that? |
Do you mind if I correct a |
Ok, there are a bunch of other errors related to Python 3-incompatibility (import stuff mostly). Should I address those too? |
I will just comment out python 3.x for now. We can address the py3 stuff in a separate PR. |
Go for it! |
I've commented out the py3 tests |
Thanks, guys! I was able to fix a ton of bugs now that I was able to run all the tests. |
Should be ready to merge once travis concurs all test failures are gone. |
OK, this seems to have been a syntax issue! I've sorted that, and am now running into geometry proposal problems:
|
I'm seeing issues with the
Here, the old system only has 22 atoms, but old atom indices 20 and above are referenced in the atom mapping. I'll insert some more assertion tests that catch exactly when this happens. |
OK, I've tried to add some tests and can't identify any issues. Hrm.... |
I've added an assertion into This seems to fail for some mutations:
Still trying to track this down. |
OK, the bug was that some of the The remaining errors are NCMC issues! |
There appear to be some geometry engine issues too:
|
The NCMC failures appear to be due to the energy becoming
|
Whoops, looks like I was setting a bunch of other random non-alchemical context parameters to zero, like
|
OK, I'm down to just geometry engine errors:
|
…ck of available torsions. Fix bug in test_elimination.
OK, it seems like this failure occurs only for ring-containing atoms:
It could be that only impropers torsions are defined for this atom. |
I'm running into a few NCMC-based
|
Whoops, spoke too soon. There's another issue now that I've fixed the alkanes test case:
|
Looks like resolving this last error is pending changes to The other issues I need to address are
I'll merge this now and address those separately. |
Fix sampler bugs, cut down NCMC testing time, add alkanes test system
I'm very confused about the switch between I would at least like to swap the names, if this is truly fixing the bug, but it would be helpful to understand what has happened here because in the future I would like to eliminate the |
I don't know the exact cause, but I think that Feel free to do some refactoring in a PR if you want! I think we have pretty good test coverage right now, though we desperately need to fix the |
This cuts down the NCMC testing time to speed up travis.