[RLlib] Add unit tests for updating episode data in base_env #17137
Conversation
mvindiola1
force-pushed
the
16683_missing_info_dictionary
branch
from
7084832 to
da423e2
Compare
| @@ -0,0 +1,143 @@ | |||
| import ray | |||
There was a problem hiding this comment.
nice! we just have to add this into rllib/BUILD to actually get this to run
There was a problem hiding this comment.
@richardliaw It has been added to rllib/BUILD and is passing
mvindiola1
force-pushed
the
16683_missing_info_dictionary
branch
from
da423e2 to
ea89d0e
Compare
|
@mvindiola1 Could you resolve the merge conflict and formatting errors? Thanks! |
mvindiola1
force-pushed
the
16683_missing_info_dictionary
branch
from
ea89d0e to
da90238
Compare
|
@juliusfrost, I just rebased and pushed. I am not sure what to do about the formatting issues. When I run format.sh it complains about two formatting issues. If I adjust those lines to try and fix the complaint then the next time I run format.sh it changes them back and complains again. 🤷 |
|
@mvindiola1 Did you install the right linting dependencies? |
…3_missing_info_dictionary # Conflicts: # rllib/BUILD
|
Just waiting for LINT tests to pass now ... |
|
@mvindiola1 ^ |
Why are these changes needed?
@sven1977, This is a follow up to issue #16683.
I had created a unit test for this issue but krfricke beat me to the PR so it did not make it. It is in this PR if you want it. It basically tests to make sure that the last_{obs,reward,done,info,action} are being set correctly for both single and multi-agent envs.
I also made one change to the base_env.py poll method. There is an extra clearing of self.last_info at the end of the method that I do not thin should be there. It is not there for any of the other last_x so I think it was put there incorrectly and should have been removed in yesterdays PR but was not.
Checks
scripts/format.shto lint the changes in this PR.