search

mila-iqia / climate-cooperation-competition

AI for Global Climate Cooperation: Modeling Global Climate Negotiations, Agreements, and Long-Term Cooperation in RICE-N. ai4climatecoop.org
Other
37 stars 27 forks source link

removed the cast reward dict keys to string, fixes the 0 reward issue #38

Closed pjwozny closed 9 months ago

pjwozny commented 9 months ago

removed the cast reward dict keys to string, fixes the 0 reward issue

Is there a certain instance where we need the reward dict keys to be string? If not, this should do the trick.

pjwozny commented 9 months ago

one extra addition to make the extraction of the env_obj a bit more flexible. if i have num_workers > 0, then my local_worker is an empty object. this change just gets it the env_obj from all workers and takes the env from the 1st. The 0th is the local_worker which is still empty. Unsure why that is the case, but this works and trains.

AndrewRWilliams commented 9 months ago

Hmmm interesting, I had initially added to string to get it to work with the new version of Ray. I'll test locally tomorrow or Tuesday, and merge when I confirm it works

pjwozny commented 9 months ago

Strange, for ref heres my ray version from my conda env: ray 2.8.1 pypi_0 pypi

AndrewRWilliams commented 9 months ago

Works for python 3.7, python 3.9 with ray==2.7.2 (this is the latest version compatible with python 3.7).