Skip to content

Experimental support for stable_baselines3#86

Merged
blumu merged 1 commit intomainfrom
wiblum/sb
Sep 24, 2022
Merged

Experimental support for stable_baselines3#86
blumu merged 1 commit intomainfrom
wiblum/sb

Conversation

@blumu
Copy link
Copy Markdown
Contributor

@blumu blumu commented Sep 22, 2022

No description provided.

@blumu blumu force-pushed the wiblum/sb branch 4 times, most recently from b715508 to db7d639 Compare September 23, 2022 20:35
- Observation fields can now be optionally padded to the shape
expected by their corresponding gym space. (Requires more memory but is
needed to train with stable-baseline agents)
- Gym wrappers to flatten the Action and Observation spaces from CyberBattleSim
- Hack CyberBattleSim environment to allow invalid moves and
return negative reward instead
- Flatten multi-dimensioned `MultiBinary` spaces
    * works with spaces.MultiBinary([list]) and spaces.MultiBinary(number)
    * working with `nodes_privilegelevel`
    * works with `leaked_credentials`
    * works with `credential_cache_matrix`
    * works with `discovered_nodes_properties`
- Add a `stable-baseline` test notebook
(Requires custom patch of stable-baseline3
from bug DLR-RM/stable-baselines3#1073)
- Fix python 3.8 warnings
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant