Fix LRU cache for ray and update 'min_atoms' option

The ray actors were failing to properly import the cython code in the lru-dict library. Kept giving the error AttributeError: type object 'LRU' has no attribute '__len__'. Switching to pylru (pure python) fixed the issues.

Also updated the builder and molecule state objects to respect the min_atoms parameter meaning no terminal state will be an available action until at least min_atoms heavy atoms are present in the molecule.

Add header to log file

Minor release to include a header line in the rewards log csv file

Update to use ray v2.3.1 and graph-env v0.2.3

Merge pull request #117 from NREL/ray_v2_3

Update to use ray v2.3 and gymnasium v0.26.3

Update to use ray v2.2.0, graph-env v0.2.0

Ray v2.2.0 uses new algorithm config objects in RLlib instead of config dicts.

manifest.in fixes

v0.0.5

manifest again

Another attempt at including preprocessor.json

What's Changed

• QED on aws by @dmdu in #113
• QED on aws by @dmdu in #114

Full Changelog: v0.0.3...v0.0.4

v0.0.3

Initial tests for pypi

Adding code ocean single-node tests

What's Changed

• initial tests getting a single-node version of stable radical working by @pstjohn in #105

Full Changelog: v0.0.1...v0.0.2

Initial release for zenodo DOI

What's Changed

• Fixing policy model visit saving by @pstjohn in #32
• WIP continuation of charles' branch by @pstjohn in #40
• working out the hallway example, currently stuck on something related… by @pstjohn in #44
• WIP, first pass at abstracting out TF functionality into a TFAlphaZeroProblem class by @pstjohn in #45
• Moving molecule game, updating molecule construction code by @pstjohn in #48
• Minimum working example for running QED using yuma and slurm on eagle by @jlaw9 in #49
• 2021 03 04 tfaz atari by @erskordi in #50
• race condition IntegrityError fix by @pstjohn in #51
• adding StateTable to cache policy inputs by @pstjohn in #53
• Examples of solving two gym envs with AZ by @davebiagioni in #54
• Check for cycles on adding new edges by @pstjohn in #55
• Refactor molecule building code by @pstjohn in #59
• Stable radical optimization merge by @jlaw9 in #52
• try/except in gdb filter by @pstjohn in #62
• Config file containing AZ, MCTS, and database parameters by @jlaw9 in #60
• initial deploy configurations by @pstjohn in #63
• fixing some issues raised by LGTM by @pstjohn in #64
• some pep8 consistency edits with yapf by @pstjohn in #65
• Fix for #68 by @pstjohn in #70
• Fix for #69 (second attempt) by @pstjohn in #72
• Gridworld benchmarking with index-based obs space. by @davebiagioni in #73
• Initial attempt of logP optimization by @jlaw9 in #66
• fixes #75 by @pstjohn in #78
• fixes #76 by @pstjohn in #79
• initial performance metrics implementation by @pstjohn in #80
• test support for python 3.9 by @pstjohn in #81
• Move to using hashlib instead of trying to manage salting in pythons … by @pstjohn in #84
• Try testing with conda-forge's tensorflow by @pstjohn in #87
• Running AZ optimization starting with known radical centers by @pstjohn in #85
• Code to generate canonical tautomers by @pstjohn in #91
• edits during meeting by @pstjohn in #92
• Updates when generating new figures for the radical paper by @pstjohn in #94
• WIP: Policy model for crystal actions by @jlaw9 in #93
• Crystal action space with example problem by @jlaw9 in #90
• archiving old local script and adding script with edited container pa… by @scottwsides in #97
• Local postgresql server for qed example by @harrysorensennrel in #98
• 2021 06 22 parametric molecule builder by @pstjohn in #95
• Use multiple Eagle nodes to train PPO for qed optimization by @davebiagioni in #100
• fix for CI failures by @pstjohn in #103
• Basic example of running RLLib on AWS by @davebiagioni in #104

New Contributors

• @jlaw9 made their first contribution in #49
• @scottwsides made their first contribution in #97
• @harrysorensennrel made their first contribution in #98

Full Changelog: https://github.com/NREL/rlmolecule/commits/v0.0.1