This repo is forked from NeuroSAT by Selsam et al. For more information on NeuroSAT, please visit: https://github.com/dselsam/neurosat/
Similar to the main neuroSAT repo, the scripts/
directory includes all the necessary scripts to get started.
- Run
sh scripts/setup.sh
to install dependencies. Note that you need python version to be in range [3.3, 3.6] in order to install tensorflow 1.4.0. Its best to use python 3.6. - Run
sh scripts/toy_gen_data.sh
to generate toy train and test data. - Run
python3 python/normalize_data.py
to normalize the labels (times) to be in range [0,1]. Note that the file names for train/test data are currently hardcoded here and need to be modified based on the file names generated in the previous step. This needs to be updated later to take in file names as an arguement. - Run
sh scripts/toy_train.sh
to train a model for a few iterations on the toy training data. - Run
sh scripts/toy_test.sh
to evaluate the trained model on the toy test data. - Run
sh scripts/toy_solve.sh
to (try to) solve the toy test problems. - Run
sh scripts/toy_pipeline.sh
to runtoy_gen_data.sh
,toy_train.sh
,toy_test.sh
, andtoy_solve.sh
in sequence. Do not use this for now, since labels are not automatically normalized yet.