HumanCompatibleAI / adversarial-policies

Find best-response to a fixed policy in multi-agent RL
MIT License
275 stars 47 forks source link

Upgrade Baselines & make MPI dependency optional #22

Closed AdamGleave closed 5 years ago

AdamGleave commented 5 years ago

Removed MPI from Dockerfile.

We do still need it for some algorithms (PPO1, GAIL, SAC) but we're not using these right now, so let's avoid it for now, hopefully some of the bugs will get fixed before when we next need to use it.

codecov[bot] commented 5 years ago

Codecov Report

Merging #22 into master will decrease coverage by <.01%. The diff coverage is 83.33%.

Impacted file tree graph

@@            Coverage Diff            @@
##           master     #22      +/-   ##
=========================================
- Coverage    58.3%   58.3%   -0.01%     
=========================================
  Files          57      58       +1     
  Lines        4895    4914      +19     
=========================================
+ Hits         2854    2865      +11     
- Misses       2041    2049       +8
Flag Coverage Δ
#aprl 10.8% <0%> (-0.05%) :arrow_down:
#modelfree 51.36% <83.33%> (+0.02%) :arrow_up:
Impacted Files Coverage Δ
src/modelfree/training/gail_dataset.py 100% <100%> (ø)
src/modelfree/multi/common.py 71.71% <100%> (ø) :arrow_up:
src/modelfree/policies/base.py 85.91% <50%> (-1.05%) :arrow_down:
src/modelfree/policies/loader.py 81.55% <60%> (-2.49%) :arrow_down:
src/modelfree/train.py 89.28% <85.71%> (-1.07%) :arrow_down:
src/modelfree/visualize/annotated_gym_compete.py 83.03% <0%> (-0.61%) :arrow_down:

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 4928326...74306ed. Read the comment docs.