[ICML 2020] DrRepair: Learning to Repair Programs from Error Messages

Overview

DrRepair: Learning to Repair Programs from Error Messages

This repo provides the source code & data of our paper: Graph-based, Self-Supervised Program Repair from Diagnostic Feedback (ICML 2020).

@InProceedings{Yasunaga20DrRepair,
  author =  {Michihiro Yasunaga and Percy Liang},
  title =   {Graph-based, Self-Supervised Program Repair from Diagnostic Feedback},
  year =    {2020},  
  booktitle =   {International Conference on Machine Learning (ICML)},  
}

Dependencies

  • GCC: Follow the SPoC requirement (https://github.com/Sumith1896/spoc)
  • Python 3.6.8 (e.g. conda create -n DrRepair python=3.6.8)
  • Python libraries
    • torch==1.0.1, numpy, tqdm, regex, joblib, pyyaml, bottle, cheroot, tensorboardX
    • clang==8.0.1 (do the following)
      conda config --add channels conda-forge
      conda install python-clang==8.0.1
      

Data

Download all the raw data -- DeepFix, SPoC, codeforce (for pretraining) -- by

./download_raw_data.sh

You can preprocess the raw data to get the program repair data by running the commands in

data/1.run-gen-err-dataset--orig-spoc.sh
data/2.run-gen-err-dataset--auto-corrupt--spoc.sh
data/3.run-gen-err-dataset--auto-corrupt--deepfix.sh

However, this takes a significant time, so for your convenience, you can download all the preprocessed data by

./download_preprocessed_data.sh

The repo structure looks like the following:

.
└─ raw_data/
   ├── codeforce_data/                  (raw programs from codeforce)
   ├── deepfix_data/                    (raw programs from deepfix)
   └── spoc_data/
       ├── spoc                              (SPoC data release)
       └── translation_preds                 (line-level code predictions from Kulal+19)

└─ data/                             
   ├── *.sh, *.py                       (preprocessing scripts)
   ├── err-data-compiler--orig-spoc/    (preprocessed, program repair data for spoc)
   ├── err-dev-compiler--for-SPoC/      (└─ dev data for spoc)
   ├── err-vocab-compiler--for-SPoC/    (└─ vocab for spoc)
   ...
   ... [similarly for deepfix and pre-training]

└─ utils/                      (utilities for code processing)

└─ model/                      (DrRepair model)

└─ evaluation/                 (to evaluate Repair model on deepfix/spoc test)
   ├── deepfix
   └── spoc
       ├── translation_preds_test/           (line-level code predictions from Kulal+19 for TestP/TestW)
       ...

Train models

Let's train program repair models. First, go to model directory. Then, run commands listed in run_deepfix.sh or run_spoc.sh. For example, if we train DrRepair ("base + graph" in the paper) on the DeepFix data, run:

name="code-compiler--2l-graph"
mkdir -p out_deepfix/${name}
python3 -u main_deepfix.py -o ${name} train \
    configs/base.yml  configs/data-deepfix/err-data-orig.yml \
    configs/model-code-compiler/2l-graph--dec-attn-all.yml

Evaluate models

We run the trained program repair model as a server. We then call this model on application tasks (DeepFix and SPoC) to evaluate the usefulness of the model.

DeepFix

1. Start server

First, go to model directory. We run a trained model (e.g. code-compiler--2l-graph) as a server by

name="SERVER--code-compiler--2l-graph"
mkdir out_deepfix/${name}
python3 -u main_deepfix.py -o ${name} server -p <port> \
    -l out_deepfix/code-compiler--2l-graph/<checkpoint> \
    configs/base.yml  configs/data-deepfix/err-data-orig.yml \
    configs/model-code-compiler/2l-graph--dec-attn-all.yml

For <port>, pick a port number (e.g. 8080) for the server. For <checkpoint>, pick a checkpoint (e.g. 150000) of the trained model. Then run ifconfig to get the IP address (e.g. 172.24.67.161) of the machine hosting this model. Concrete examples are provided in the second half of model/run_deepfix.sh.

2. Run model on DeepFix test

Go to evaluation/deepfix directory. First prepare:

repo_root="../../../.."
program_data_root=${repo_root}"/raw_data/deepfix_data"
test_split_root=${repo_root}"/data/err-data-compiler--auto-corrupt--orig-deepfix/bin4"

To run the trained model on the DeepFix test examples, do

name="code-compiler--2l-graph"
mkdir -p out/${name}/log
cd out/${name}

for entry in ${test_split_root}/*
do
  probid=`basename $entry`
  python3 -u ../../test_deepfix.py \
  --input-code-dir ${program_data_root}/${probid}/erroneous \
  --repairer-server  http://<IP>:<port>/pred
done

where you plug the IP address and port number into <IP> and <port>. After this completes, you can get the test accuracy by

python3 -u ../../collate_deepfix.py

Concrete examples are provided in evaluation/run_test_deepfix.sh.

SPoC

1. Start server

First, go to model directory. We run a trained model (e.g. code-compiler--2l-graph--finetune) as a server by

name="SERVER--code-compiler--2l-graph--finetune"
mkdir out_spoc/${name}
python3 -u main_spoc.py -o ${name} server -p <port> \
    -l out_spoc/code-compiler--2l-graph--finetune/<checkpoint> \
    configs/base.yml  configs/data-spoc/err-data-orig.yml \
    configs/model-code-compiler/2l-graph--dec-attn-all.yml

Similar to DeepFix, pick a port number and a checkpoint, and get the IP address. Concrete examples are provided in the second half of model/run_spoc.sh.

2. Run model on SPoC test

Go to evaluation/spoc directory. First prepare:

repo_root="../../../.."

To run the trained model on all the programs in SPoC TestW, do

name="code-compiler--2l-graph--finetune"

INPUT=translation_preds_test/testw    #change to testp if you want to evaluate on testp
N=$(tail -n+2 ${INPUT}.tsv | cut -f 3-6 | uniq | wc -l)  # Count the number of programs
interval=10

mkdir -p out_testw/${name}/log        #change to testp if you want to evaluate on testp
cd out_testw/${name}                  #change to testp if you want to evaluate on testp

i=1
while [[ $i -le $N ]]; do
  python -u ../../test_spoc.py -p 100 \
  --compile-budget 100 --n-parallel ${interval} \
  --repairer-server  http://<IP>:<port>/pred \
  ../../${INPUT} $i
  i=$(($i + ${interval}))
done

where you plug the IP address and port number into <IP> and <port>. After this completes, you can get the test accuracy by

python3 -u ../../collate_spoc.py

Concrete examples are provided in evaluation/run_test_spoc.sh.

Acknowledgment

The original DeepFix and SPoC data used in this work come from the following papers:

DeepFix: Fixing common C language errors by deep learning. Rahul Gupta, Soham Pal, Aditya Kanade, Shirish Shevade. AAAI 2017.
SPoC: Search-based Pseudocode to Code. Sumith Kulal, Panupong Pasupat, Kartik Chandra, Mina Lee, Oded Padon, Alex Aiken and Percy Liang. NeurIPS 2019.
Owner
Michihiro Yasunaga
PhD Student in Computer Science
Michihiro Yasunaga
PantheonRL is a package for training and testing multi-agent reinforcement learning environments.

PantheonRL is a package for training and testing multi-agent reinforcement learning environments. PantheonRL supports cross-play, fine-tuning, ad-hoc coordination, and more.

Stanford Intelligent and Interactive Autonomous Systems Group 57 Dec 28, 2022
The Python ensemble sampling toolkit for affine-invariant MCMC

emcee The Python ensemble sampling toolkit for affine-invariant MCMC emcee is a stable, well tested Python implementation of the affine-invariant ense

Dan Foreman-Mackey 1.3k Dec 31, 2022
Python implementation of Wu et al (2018)'s registration fusion

reg-fusion Projection of a central sulcus probability map using the RF-ANTs approach (right hemisphere shown). This is a Python implementation of Wu e

Dan Gale 26 Nov 12, 2021
This repository contains code to train and render Mixture of Volumetric Primitives (MVP) models

Mixture of Volumetric Primitives -- Training and Evaluation This repository contains code to train and render Mixture of Volumetric Primitives (MVP) m

Meta Research 125 Dec 29, 2022
Conservative Q Learning for Offline Reinforcement Reinforcement Learning in JAX

CQL-JAX This repository implements Conservative Q Learning for Offline Reinforcement Reinforcement Learning in JAX (FLAX). Implementation is built on

Karush Suri 8 Nov 07, 2022
Official re-implementation of the Calibrated Adversarial Refinement model described in the paper Calibrated Adversarial Refinement for Stochastic Semantic Segmentation

Official re-implementation of the Calibrated Adversarial Refinement model described in the paper Calibrated Adversarial Refinement for Stochastic Semantic Segmentation

Elias Kassapis 31 Nov 22, 2022
Code for our ALiBi method for transformer language models.

Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolation This repository contains the code and models for our paper Tra

Ofir Press 211 Dec 31, 2022
Degree-Quant: Quantization-Aware Training for Graph Neural Networks.

Degree-Quant This repo provides a clean re-implementation of the code associated with the paper Degree-Quant: Quantization-Aware Training for Graph Ne

35 Oct 07, 2022
Official Pytorch implementation of "Unbiased Classification Through Bias-Contrastive and Bias-Balanced Learning (NeurIPS 2021)

Unbiased Classification Through Bias-Contrastive and Bias-Balanced Learning (NeurIPS 2021) Official Pytorch implementation of Unbiased Classification

Youngkyu 17 Jan 01, 2023
This program generates a random 12 digit/character password (upper and lowercase) and stores it in a file along with your username and app/website.

PasswordGeneratorAndVault This program generates a random 12 digit/character password (upper and lowercase) and stores it in a file along with your us

Chris 1 Feb 26, 2022
An implementation of "Optimal Textures: Fast and Robust Texture Synthesis and Style Transfer through Optimal Transport"

Optex An implementation of Optimal Textures: Fast and Robust Texture Synthesis and Style Transfer through Optimal Transport for TU Delft CS4240. You c

Hans Brouwer 33 Jan 05, 2023
Tensorflow Implementation for "Pre-trained Deep Convolution Neural Network Model With Attention for Speech Emotion Recognition"

Tensorflow Implementation for "Pre-trained Deep Convolution Neural Network Model With Attention for Speech Emotion Recognition" Pre-trained Deep Convo

Ankush Malaker 5 Nov 11, 2022
Toontown House CT Edition

Toontown House: Classic Toontown House Classic source that should just work. ❓ W

Open Source Toontown Servers 5 Jan 09, 2022
An essential implementation of BYOL in PyTorch + PyTorch Lightning

Essential BYOL A simple and complete implementation of Bootstrap your own latent: A new approach to self-supervised Learning in PyTorch + PyTorch Ligh

Enrico Fini 48 Sep 27, 2022
This program was designed to detect whether someone is wearing a facemask through a live video stream.

This program was designed to detect whether someone is wearing a facemask through a live video stream. A custom lightweight CNN trained with TensorFlow on a public dataset provided by Kaggle is used

0 Apr 02, 2022
Self-training with Weak Supervision (NAACL 2021)

This repo holds the code for our weak supervision framework, ASTRA, described in our NAACL 2021 paper: "Self-Training with Weak Supervision"

Microsoft 148 Nov 20, 2022
Semi-supervised Adversarial Learning to Generate Photorealistic Face Images of New Identities from 3D Morphable Model

Semi-supervised Adversarial Learning to Generate Photorealistic Face Images of New Identities from 3D Morphable Model Baris Gecer 1, Binod Bhattarai 1

Baris Gecer 190 Dec 29, 2022
TiP-Adapter: Training-free CLIP-Adapter for Better Vision-Language Modeling

TiP-Adapter: Training-free CLIP-Adapter for Better Vision-Language Modeling This is the official code release for the paper 'TiP-Adapter: Training-fre

peng gao 189 Jan 04, 2023
Label-Free Model Evaluation with Semi-Structured Dataset Representations

Label-Free Model Evaluation with Semi-Structured Dataset Representations Prerequisites This code uses the following libraries Python 3.7 NumPy PyTorch

8 Oct 06, 2022
A super lightweight Lagrangian model for calculating millions of trajectories using ERA5 data

Easy-ERA5-Trck Easy-ERA5-Trck Galleries Install Usage Repository Structure Module Files Version iteration Easy-ERA5-Trck is a super lightweight Lagran

Zhenning Li 26 Nov 19, 2022