Use Sage 2.2.0 RC1 as default

[mattwthompson]

#306

[jthorton]

Thanks @mattwthompson it would be great to make this change!

[codecov]

Codecov Report

Merging #307 (f1ba30b) into main (f7cd2a6) will increase coverage by 0.05%.
The diff coverage is n/a.

Additional details and impacted files

[mattwthompson]

No idea why the OpenEye integration tests are failing ...

[jthorton]

Oh no looks like some strange interaction with fitting on top of this force field. We did see an error like this before but that was fixed by using the new geometric #243.

[mattwthompson]

Any other ideas where to start? The main thing that stands out to me is the toolkit difference, but I doubt that either

• Omega would give an ethane conformer so bad it can't be optimized or
• ELF10 gives substantively different charges such that the optimization should crash

[mattwthompson]

I don't know how to isolate this behavior locally, and there have been some other quiet issues bumping OpenEye Toolkits from 2022 to 2023, so I'm just going to constrain down to what we know worked before

[j-wags]

@jthorton Any other ideas? Matt and I just looked at this and are stumped. It looks like we're using the latest geometric too.

[mattwthompson]

A more useful error:

/Users/mattthompson/micromamba/envs/bespokefit-test/lib/python3.11/site-packages/forcebalance/smirnoffio.py:422: InterchangeDeprecationWarning: The `handlers` attribute is deprecated. Use `collections` instead.
  if 'VirtualSites' in interchange.handlers:
Traceback (most recent call last):
  File "/Users/mattthompson/micromamba/envs/bespokefit-test/bin/ForceBalance.py", line 45, in Run_ForceBalance
    optimizer.Run()
  File "/Users/mattthompson/micromamba/envs/bespokefit-test/lib/python3.11/site-packages/forcebalance/optimizer.py", line 322, in Run
    xk = self.OptTab[self.jobtype]()
         ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/mattthompson/micromamba/envs/bespokefit-test/lib/python3.11/site-packages/forcebalance/optimizer.py", line 951, in NewtonRaphson
    return self.MainOptimizer(b_BFGS=0)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/mattthompson/micromamba/envs/bespokefit-test/lib/python3.11/site-packages/forcebalance/optimizer.py", line 676, in MainOptimizer
    dx, dX_expect, bump = self.step(xk, data, trust)
                          ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/mattthompson/micromamba/envs/bespokefit-test/lib/python3.11/site-packages/forcebalance/optimizer.py", line 929, in step
    Result = optimize.brent(search_fun,brack=(LOpt,LOpt*4),tol=self.search_tol,full_output=1)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/mattthompson/micromamba/envs/bespokefit-test/lib/python3.11/site-packages/scipy/optimize/_optimize.py", line 2730, in brent
    res = _minimize_scalar_brent(func, brack, args, **options)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/mattthompson/micromamba/envs/bespokefit-test/lib/python3.11/site-packages/scipy/optimize/_optimize.py", line 2767, in _minimize_scalar_brent
    brent.optimize()
  File "/Users/mattthompson/micromamba/envs/bespokefit-test/lib/python3.11/site-packages/scipy/optimize/_optimize.py", line 2537, in optimize
    xa, xb, xc, fa, fb, fc, funcalls = self.get_bracket_info()
                                       ^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/mattthompson/micromamba/envs/bespokefit-test/lib/python3.11/site-packages/scipy/optimize/_optimize.py", line 2506, in get_bracket_info
    xa, xb, xc, fa, fb, fc, funcalls = bracket(func, xa=brack[0],
                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/mattthompson/micromamba/envs/bespokefit-test/lib/python3.11/site-packages/scipy/optimize/_optimize.py", line 3136, in bracket
    raise e
scipy.optimize._optimize.BracketError: The algorithm terminated without finding a valid bracket. Consider trying different initial points.

[j-wags]

I don't think we should merge this in a form where a RC is the default FF. It's fine if we use it as part of the debugging process but we shouldn't put an RC into production.

[mattwthompson]

Right, but this should be ready to go on release

[jthorton]

Well this is very confusing as the new RC passes all the tests.

[mattwthompson]

Won't work until conda-forge/openff-forcefields-feedstock#25 is through

[j-wags]

The new FF package should be available now. Fixed a little typo and kicked CI.

[mattwthompson]

This has one approving review, which is fairly old but in the spirit of its current state. Everything is green, so I'm merging