https://github.com/sri-csl/coproof

https://github.com/sri-csl/coproof

Last synced: 15 days ago
JSON representation

• Host: GitHub
• URL: https://github.com/sri-csl/coproof
• Owner: SRI-CSL
• License: gpl-2.0
• Created: 2023-07-24T22:42:18.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2023-10-23T16:30:36.000Z (about 1 year ago)
• Last Synced: 2024-11-06T04:36:46.664Z (2 months ago)
• Language: Jupyter Notebook
• Size: 45.9 MB
• Stars: 1
• Watchers: 17
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# CoProver

CoProver is a proof recommendation system for proof assistants trained on PVS and Imandra

## Installation

CoProver is intended to run on Python 3.9+.  Once an environment has been selected, install the dependencies via `setup.py`

```

cd coproof

pip install -e .

```

## Tasks

### Training and running the PVS Command Predictor experiments

To train and evaluate the PVS command prediction, execute this module,

```

python -m coprover.cmdpred.train_t5

```

Note this uses PyTorch Lightning to do DDP training across several GPUs.  You may need to adjust the `num_gpus` argument to match the number of GPUs you have on your system, or use 0 for software only.

The baselines for comparison can be executed via,

```

python -m coprover.cmdpred.run_baselines

```

### Training and running the PVS Lemma Retrieval experiments

To first train the underlying Sentence BERT model, 

```

python -m coprover.lemmaret.bin.train_sbert

```

### Generating JSON formatted data from PVS JSON proof traces

python -m coprover.feats.featurize_cmdpred