*Refactoring is undergoing.

BlendRL: A Framework for Merging Symbolic and Neural Policies

Hikaru Shindo, Quentin Delfosse, Devendra Singh Dhami, Kristian Kersting

We propose a framework that jointly learns symbolic and neural policies for reinforcement learning.

Quickstart

Install nsfr and nudge.

Training script:

python train_blenderl.py --env-name seaquest --joint-training --num-steps 128 --num-envs 5 --gamma 0.99

• --joint-training: train neural and logic modules jointly
• --num-steps: the number of steps for policy rollout
• --num-envs: the number of environments to train agents
• --gamma: the discount factor for future rewards

Play script:

python play_gui.py

Note that a checkpoint is required to run the play script.

How to Use

Hyperparameters

The hyperparameters are configured inside in/config/default.yaml which is loaded as default. You can specify a different configuration by providing the corresponding YAML file path as an argument, e.g., python train.py in/config/my_config.yaml. A description of all hyperparameters can be found in train.py.

The Logic

Inside in/envs/[env_name]/logic/[ruleset_name]/, you find the logic rules that are used as a starting point for training. You can change them or create new rule sets. The ruleset to use is specified with the hyperparam rules.

Install Locally

If you want to use NUDGE within other projects, you can install NUDGE locally as follows:

1. Inside nsfr/ run

   python setup.py develop

2. Inside nudge/ run

   python setup.py develop

Ohter dependencies

1. Install packages by pip install -r requirements.txt

2. PyG and torch-scatter for neumann Install PyG and torch-scatter packages for neumann reasoner. See the installation guide. These should be consistent in terms of ther versions, e.g.

   pip install torch==1.12.0+cu116 torchvision==0.13.0+cu116 -f https://download.pytorch.org/whl/torch_stable.html
   pip install torch_geometric
   pip install pyg_lib torch_scatter torch_sparse -f https://data.pyg.org/whl/torch-1.12.0+cu116.html
   

Playing

TODO

Example to play with a trained ppo agent

python3 play.py -s 0 -alg ppo -m getout -env getout  

How to Set up New Environments

You add a new environment inside in/envs/[new_env_name]/. There, you need to define a NudgeEnv class that wraps the original environment in order to do

• logic state extraction: translates raw env states into logic representations
• valuation: Each relation (like closeby) has a corresponding valuation function which maps the (logic) game state to a probability that the relation is true. Each valuation function is defined as a simple Python function. The function's name must match the name of the corresponding relation.
• action mapping: action-predicates predicted by the agent need to be mapped to the actual env actions

See the freeway env to see how it is done.