Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

https://github.com/facebookresearch/CodeGen

Reference implementation of code generation projects from Facebook AI Research. General toolkit to apply machine learning to code, from dataset creation to model training and evaluation. Comes with pretrained models.
https://github.com/facebookresearch/CodeGen

Last synced: 3 months ago
JSON representation

Reference implementation of code generation projects from Facebook AI Research. General toolkit to apply machine learning to code, from dataset creation to model training and evaluation. Comes with pretrained models.

Lists

README

        

This repository is a toolkit to do machine learning for programming languages. It implements tokenization, dataset preprocessing, model training and model evaluation.

We provide reference implementations of the following papers:
- [TransCoder: Unsupervised Translation of Programming Languages](https://arxiv.org/pdf/2006.03511.pdf) (2020)
- [DOBF: A Deobfuscation Pre-Training Objective for Programming Languages](https://arxiv.org/pdf/2102.07492.pdf) (2021)
- [TransCoder-ST: Leveraging Automated Unit Tests for Unsupervised Code Translation](https://arxiv.org/pdf/2110.06773.pdf) (2021)
- [TransCoder-IR: Code Translation with Compiler Representations](https://arxiv.org/pdf/2207.03578.pdf) (2022)

We also provide pre-trained models for language modeling, translation and deobfuscation.

You can find some documentation for each projects in the docs folder:
- [TransCoder](docs/transcoder.md).
- [DOBF](docs/dobf.md)
- [TransCoder-ST](docs/TransCoder-ST.md)
- [TransCoder-IR](docs/TransCoder-IR.md)

## Dependencies
Run [install_env.sh](install_env.sh).
We use black code formatter.

## Data
### Source code processors

This repository contains [programming languages processors](codegen_sources/preprocessing/lang_processors/lang_processor.py) for C++, Java and Python. These processors include:
- tokenization and detokenization
- obfuscation
- function extractions

These processors are based on [TreeSitter](https://tree-sitter.github.io/tree-sitter/) parsers. As these parsers are available in more than 30 programming languages, one can easily create a new programming language processor.

Example of code tokenization:

```python
from codegen_sources.preprocessing.lang_processors.java_processor import JavaProcessor

java_code = r"""class HelloWorld {
public static void main(String[] args) {
System.out.println("Hello, World!");
}
}"""
java_processor = JavaProcessor(root_folder="")
tokenized_java_code = java_processor.tokenize_code(java_code)
print(tokenized_java_code)
```

### BPE
This repository provides wrappers for [fast BPE](codegen_sources/preprocessing/bpe_modes/fast_bpe_mode.py) and [Roberta BPE](codegen_sources/preprocessing/bpe_modes/roberta_bpe_mode.py) at file level.

### Dataset Preprocessing

This repository contains a [pipeline](codegen_sources/preprocessing/preprocess.py) to create programming languages datasets. Now it supports [four datasets modes](codegen_sources/preprocessing/dataset_modes):
- Monolingual (ex: Java source code)
- Monolingual Functions (ex: Java functions)
- Monolingual Obfuscated (ex: Obfuscated Java source code.)
- Monolingual Obfuscated Functions (ex: Obfuscated Java functions)

First, download C++ / Java / Python source code from [Google BigQuery](https://cloud.google.com/blog/products/gcp/github-on-bigquery-analyze-all-the-open-source-code). To run our preprocessing pipeline, you need to donwload the raw source code on your machine in a JSON format. A sample of it is given [here](data/test_dataset).

The pipeline does the following:
- Source code extraction from json (`.json.gz`) and tokenization (`.tok`)
- Train BPE codes and vocab
- Apply BPE (`.bpe`)
- Binarization (`.pth`)
- Symlink folder with appropriate file names for `.pth` (XLM-syml). To be given as `data_path` argument for training.

To run the pipeline :

```bash
python -m codegen_sources.preprocessing.preprocess \
\ # folder containing json.gz
--langs java cpp python \ # languages to process
--mode monolingual_functions \ # dataset mode
--bpe_mode=fast \ # BPE mode. by default it is fast. can be roberta
--local=True \ # Run on your local machine if True. If False run on a cluster (requires submitit setup)
--train_splits=1 # Number of trainings splits
```
If you give several languages, the BPE codes and vocab will be learned commonly on these languages , so that you will have a common vocabulary to train one model for several languages. If you do not want that, launch the pipeline on every language separatly. [These tests](codegen_sources/preprocessing/tests/pipeline/test_pipeline.py) test the pipeline on different modes. It will give you an overview of the possible options.

Also, we provide the BPE codes and vocabulary [here](data/bpe/cpp-java-python). These are the codes and vocabulary used for TransCoder and DOBF. They were learned on concatenated C++, Java, and Python data. If you want to use them instead of learning new ones, give the corresponding paths as ```fastbpe_code_path``` and ```fastbpe_vocab_path``` arguments.

In TransCoder and DOBF readmes, we provide the commands to preprocess the respective datasets.

## Model

### Overview
In this repository, we provide [code](codegen_sources/model) to [train](codegen_sources/model/train.py) transformer-based models (code based on [XLM repository](https://github.com/facebookresearch/XLM)). The available training tasks are the following:
- Masked Language Model (MLM)
- Causal Language Model (CLM)
- Supervised Machine translation (MT)
- Classification
- Deobfuscation = DOBF
- Unsupervised Machine translation = TransCoder (Denoising auto encoding AE + Back Translation BT)

We [evaluate](codegen_sources/model/src/evaluation/evaluator.py) our models with metrics adapted to each task (e.g. computation accuracy and BLEU score for TransCoder, subtoken score for Deobfuscation).

Also, we provide [wrappers](codegen_sources/wrappers) to fine-tune and evaluate our models on [CodeXGLUE](https://arxiv.org/pdf/2102.04664.pdf) benchmark.

### Download models
You can download the following models:
- [MLM](docs/dobf.md#pre-trained-models)
- [TransCoder](docs/transcoder.md#pre-trained-models). Use it to translate some code [here](codegen_sources/model/translate.py).
- [DOBF](docs/dobf.md#pre-trained-models). Use it to deobfuscate some code [here](codegen_sources/model/deobfuscate.py).

### Re train specific models

To have details on how to retrain specific models, please refer to the README specific to each model.
- [TransCoder README](docs/transcoder.md).
- [DOBF README](docs/dobf.md)

## References

### TransCoder model (NeurIPS 2020)

[1] B. Roziere*, M.A. Lachaux*, L. Chanussot, G. Lample [Unsupervised Translation of Programming Languages](https://research.fb.com/wp-content/uploads/2020/11/Unsupervised-Translation-of-Programming-Languages.pdf).

```
@article{roziere2020unsupervised,
title={Unsupervised translation of programming languages},
author={Roziere, Baptiste and Lachaux, Marie-Anne and Chanussot, Lowik and Lample, Guillaume},
journal={Advances in Neural Information Processing Systems},
volume={33},
year={2020}
}
```

### DOBF

[2] B. Roziere*, M.A. Lachaux*, M. Szafraniec , G. Lample [DOBF: A Deobfuscation Pre-Training Objective for Programming Languages](https://arxiv.org/abs/2102.07492).

```
@article{roziere2021dobf,
title={{DOBF}: A Deobfuscation Pre-Training Objective for Programming Languages},
author={Roziere, Baptiste and Lachaux, Marie-Anne and Szafraniec, Marc and Lample, Guillaume},
journal={arXiv preprint arXiv:2102.07492},
year={2021}
}
```

### TransCoder-ST
[3] B. Roziere, J.M. Zhang, F. Charton, M. Harman, G. Synnaeve, G. Lample [Leveraging Automated Unit Tests for Unsupervised Code Translation](https://arxiv.org/pdf/2110.06773.pdf).

```
@article{roziere2021leveraging,
title={Leveraging Automated Unit Tests for Unsupervised Code Translation},
author={Roziere, Baptiste and Zhang, Jie M and Charton, Francois and Harman, Mark and Synnaeve, Gabriel and Lample, Guillaume},
journal={ICLR},
year={2022}
}
```

### TransCoder-IR
```
@article{szafraniec2022code,
title={Code translation with Compiler Representations},
author={Szafraniec, Marc and Roziere, Baptiste and Charton, Hugh Leather Francois and Labatut, Patrick and Synnaeve, Gabriel},
journal={ICLR},
year={2023}
}
```

\* Equal Contribution

## License
The validation and test parallel datasets from GeeksForGeeks, and the evaluation scripts under [data/transcoder_evaluation_gfg](data/transcoder_evaluation_gfg) are released under the Creative Commons Attribution-ShareAlike 2.0 license. See https://creativecommons.org/licenses/by-sa/2.0/ for more information.

The rest of the `CodeGen` repository is under the MIT license. See [LICENSE](LICENSE) for more details.