This PR is (unfortunately) fairly massive, but is mostly renaming things, or removing things that aren't needed anymore.
Updated README.md
Added docs/contributing.md and docs/getting_started.md
Added (a link to) an introductory notebook, runnable on colab
Renamed FlatRewards into ObjectProperties, as well as related function names
Renamed RewardScalar into either LogScalar or LinScalar, as well as changed the function signature of functions accordingly
Removed all passing around of rng instances, replaced by get_worker_rng to remove any confusion around worker seeding
Similarly, dev should now be acquired via get_worker_device
None-proofed cond_info in models. There should now be a sensible default behavior. Eventually we should properly make sure the code support unconditional tasks
Made cfg.git_hash computational conditional on being in a git repo
Renamed generic functions and methods working with objects to obj[s] rather than mol[s]
Made terminate close log files (to avoid issues when a single process runs multiple trials)
Added a read_all_results function in sqlite_log.py. This is probably the fastest way to load log data
This PR is (unfortunately) fairly massive, but is mostly renaming things, or removing things that aren't needed anymore.
README.md
docs/contributing.md
anddocs/getting_started.md
FlatRewards
intoObjectProperties
, as well as related function namesRewardScalar
into eitherLogScalar
orLinScalar
, as well as changed the function signature of functions accordinglyrng
instances, replaced byget_worker_rng
to remove any confusion around worker seedingdev
should now be acquired viaget_worker_device
None
-proofedcond_info
in models. There should now be a sensible default behavior. Eventually we should properly make sure the code support unconditional taskscfg.git_hash
computational conditional on being in a git repoobj[s]
rather thanmol[s]
terminate
close log files (to avoid issues when a single process runs multiple trials)read_all_results
function insqlite_log.py
. This is probably the fastest way to load log datamypy
issues.