Dual Algorithmic Reasoning

Official code for the paper Dual Algorithmic Reasoning

Dependencies

reqs/* contains requirements files for linux-cpu, macos-cpu and linux-gpu

Usage

The two main files build_data.py and run.py have a Typer CLI. Refer to each script's commands by running

python <filename>.py --help

Data

python build_data.py [PARAMS] creates training/validation/test data and should be the first command you run.

Train/Test

python run.py [COMMAND] [PARAMS] runs either the model selection (valid) or the risk assessment (test).

Structure

• config/* contains configurations for data generation, model hyperparameters (included the optimal ones) and general configs.
• config/exp/* contains the YAML files for reproducing the experiments shown in the paper (use the *.one.yaml versions).
• nn/* contains all the model implementations.