Elamur Research⌕
☰
home / papers / 2022 / Compilable Neural Code Generation with Compiler Feedback

Compilable Neural Code Generation with Compiler Feedback

Paper · 2022 · Xin Wang, Yasheng Wang, Yao Wan et al.

Summary

Neural code generators, including pre-trained models such as CodeGPT, PLBART and CodeT5, treat programs as ordinary text and rarely check whether the output compiles. Broken code costs programmers time and lowers their trust in such tools. The authors propose COMPCODER, which uses compiler feedback in a three-stage pipeline:

  1. CodeGPT is fine-tuned on the target task with a standard cross-entropy objective.
  2. The model is refined with reinforcement learning, using compilation success as the reward.
  3. A discriminator is trained jointly with the generator to judge whether candidate outputs compile, which pushes the generator to fix its own errors.

Stages 2 and 3 alternate during training. The method is tested on Python code completion and on text-to-code generation. Against plain CodeGPT, the compilation success rate rises from 44.18 to 89.18 on average for completion and from 70.3 to 96.2 for text-to-code. This shows that a binary compiler signal is a cheap and effective training target. The authors note that compiling does not mean the code is correct, and they plan to use unit tests in future work.

Key points

  • Generator and RL setup: The generator is CodeGPT, a GPT-2 model further pre-trained on code. The RL stage uses PPO2 with a reward of +1.0 for compilable output and −1.0 otherwise.
  • Preventing reward exploitation: The compiler can be fooled by trivial outputs, such as strings of whitespace or near-empty completions. To stop the policy from drifting too far, a KL-divergence penalty ties it to the fine-tuned reference model.
  • Two uses of compiler signals: The signal serves as an RL reward for the generator. It also trains a compilability discriminator that is learned jointly with the generator.
  • Main results versus CodeGPT:
    • Code completion: average compilation rate rises from 44.18 to 89.18.
    • Text-to-code generation: compilation rate rises from 70.3 to 96.2.
  • Motivation from prior work: Earlier work (Chen et al., 2021) found that 67%–97% of patches from leading deep-learning models fail to compile.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🏢Huawei TechnologiesShenzhen · China105 works · 15,705 citations🎓Huazhong University of Science and TechnologyWuhan · China64 works · 8,427 citations🎓Wuhan UniversityWuhan · China52 works · 1,456 citations🎓Yunnan UniversityKunming · China21 works · 103 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 19,503 worksArtificial Intelligence · 18,805 worksArts and Humanities · 17,959 worksLanguage and Linguistics · 17,789 worksInformation Systems · 1,888 worksComputer Science Applications · 184 works

🧭 Topics 1Research subjects and languages the work is about.Code & programming languages · 194 works

Cite

Cite as Xin Wang, Yasheng Wang, Yao Wan, Fei Mi, Yitong Li, Pingyi Zhou, Jin Liu, Hao Wu, Xin Jiang and Qun Liu. 2022. Compilable Neural Code Generation with Compiler Feedback. Findings of the Association for Computational Linguistics: ACL 2022.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2022.findings-acl.2 · arXiv 2203.05132

@inproceedings{wang2022compilable,
  title = {{Compilable Neural Code Generation with Compiler Feedback}},
  author = {Xin Wang, Yasheng Wang, Yao Wan, Fei Mi, Yitong Li, Pingyi Zhou, Jin Liu, Hao Wu, Xin Jiang and Qun Liu},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2022},
  year = {2022},
  doi = {10.18653/v1/2022.findings-acl.2},
  eprint = {2203.05132},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.18653/v1/2022.findings-acl.2}
}

References

36 references, 11 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Wasi Uddin Ahmad, Saikat Chakraborty, Baishakhi Ray, and Kai-Wei Chang. 2021. Unified pre-training for program understanding and generation. In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT 20…
  2. Tomasz Korbak, Hady ElSahar, Marc Dymetman, and Germán Kruszewski. 2021. Energy-based models for code generation under compilability constraints. arXiv preprint arXiv: 2106.04985. — [arXiv:2106.04985](https://arxiv.org/abs/2106.04985) find
  3. Xinyun Chen, Chang Liu, and Dawn Song. 2018. Tree-to-tree neural networks for program translation. In Advances in Neural Information Processing Systems 31: Annual Conference on Neural Information Processing Systems 2018, NeurIPS 2018, December 3-8, 2018, Montréal, Canada, pages 2552–2562. find
  4. Sumith Kulal, Panupong Pasupat, Kartik Chandra, Mina Lee, Oded Padon, Alex Aiken, and Percy Liang. 2019. Spoc: Search-based pseudocode to code. In Advances in Neural Information Processing Systems 32: Annual Conference on Neural Information Processing Systems 2019, NeurIPS 2019, December 8-14, 2019,…
  5. Yen-Chun Chen and Mohit Bansal. 2018. Fast abstractive summarization with reinforce-selected sentence rewriting. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics, ACL 2018, Melbourne, Australia, July 15-20, 2018, Volume 1: Long Papers, pages 675–686. Associa…
  6. Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Veselin Stoyanov, and Luke Zettlemoyer. 2020. BART: denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension. In Proceedings of the 58th Annual Meeting of the…
  7. Zimin Chen, Steve Kommrusch, Michele Tufano, Louis-Noël Pouchet, Denys Poshyvanyk, and Martin Monperrus. 2021. find
  8. Sequencer: Sequence-to-sequence Jian Li, Yue Wang, Michael R. Lyu, and Irwin King. 2018. Code completion with neural attention and pointer networks. In Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence, IJCAI 2018, July 13-19, 2018, Stockholm, Sweden, pages… find
  9. Alec Radford and Karthik Narasimhan. 2018. Improving language understanding by generative pre-training. find
  10. Alec Radford, Jeff Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. 2019. Language models are unsupervised multitask learners. find
  11. Fang Liu, Ge Li, Bolin Wei, Xin Xia, Zhiyi Fu, and Zhi Jin. 2020a. A self-attentional neural architecture for code completion with multi-task learning. In ICPC ’20: 28th International Conference on Program Comprehension, Seoul, Republic of Korea, July 13-15, 2020, pages 37–47. ACM. find
  12. Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. 2020. Exploring the limits of transfer learning with a unified text-to-text transformer. J. Mach. Learn. Res., 21:140:1–140:67.
  13. Marc’Aurelio Ranzato, Sumit Chopra, Michael Auli, and Wojciech Zaremba. 2016. Sequence level training with recurrent neural networks. In 4th International Conference on Learning Representations, ICLR 2016, San Juan, Puerto Rico, May 2-4, 2016, Conference Track Proceedings.
  14. Fang Liu, Ge Li, Yunfei Zhao, and Zhi Jin. 2020b. Multi-task learning based pre-trained language model for code completion. In 35th IEEE/ACM International Conference on Automated Software Engineering, ASE 2020, Melbourne, Australia, September 21-25, 2020, pages 473–485. IEEE. find
  15. John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. 2017. Proximal policy optimization algorithms. arXiv preprint arXiv: 1707.06347. find
  16. Shuai Lu, Daya Guo, Shuo Ren, Junjie Huang, Alexey Svyatkovskiy, Ambrosio Blanco, Colin B. Clement, Dawn Drain, Daxin Jiang, Duyu Tang, Ge Li, Lidong Zhou, Linjun Shou, Long Zhou, Michele Tufano, Ming Gong, Ming Zhou, Nan Duan, Neel Sundaresan, Shao Kun Deng, Shengyu Fu, and Shujie Liu. 2021. Codexg… find
  17. Mike Schuster and Kuldip K. Paliwal. 1997. Bidirectional recurrent neural networks. IEEE Trans. Signal Process., 45(11):2673–2681. find
  18. Richard S. Sutton and Andrew G. Barto. 1998. Introduction to reinforcement learning. find
  19. Thang Luong, Hieu Pham, and Christopher D. Manning. 2015. Effective approaches to attention-based neural machine translation. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, EMNLP 2015, Lisbon, Portugal, September 17-21, 2015, pages 1412–1421. The Associati…
  20. Alexey Svyatkovskiy, Shao Kun Deng, Shengyu Fu, and Neel Sundaresan. 2020. Intellicode compose: code generation using transformer. In ESEC/FSE ’20: 28th ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering, Virtual Event, USA, November 8-13, 202… find
  21. Maxwell I. Nye, Luke B. Hewitt, Joshua B. Tenenbaum, and Armando Solar-Lezama. 2019. Learning to infer program sketches. In Proceedings of the 36th International Conference on Machine Learning, ICML 2019, 9-15 June 2019, Long Beach, California, USA, volume 97 of Proceedings of Machine Learning Resea…
  22. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. In Advances in Neural Information Processing Systems 30: Annual Conference on Neural Information Processing Systems 2017, December 4-9, 2017,…
  23. Md. Rizwan Parvez, Wasi Uddin Ahmad, Saikat Chakraborty, Baishakhi Ray, and Kai-Wei Chang. 2021. Retrieval augmented code generation and summarization. In Findings of the Association for Computational Linguistics: EMNLP 2021, Virtual Event / Punta Cana, Dominican Republic, 16-20 November, 2021, page…
  24. Yao Wan, Zhou Zhao, Min Yang, Guandong Xu, Haochao Ying, Jian Wu, and Philip S. Yu. 2018. Improving automatic source code summarization via deep reinforcement learning. In Proceedings of the 33rd ACM/IEEE International Conference on Automated Software Engineering, ASE 2018, Montpellier, France, Sept… find
  25. Wenhua Wang, Yuqun Zhang, Yulei Sui, Yao Wan, Zhou Zhao, Jian Wu, Philip S. Yu, and Guandong Xu. 2022. Reinforcement-learning-guided source code summarization using hierarchical attention. IEEE Transactions on Software Engineering, 48(1):102–119. find
  26. Maxim Rabinovich, Mitchell Stern, and Dan Klein. 2017. Abstract syntax networks for code generation and semantic parsing. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, ACL 2017, Vancouver, Canada, July 30 -August 4, Volume 1: Long Papers, pages 1139–1149… find
  27. Xin Wang, Yasheng Wang, Fei Mi, Pingyi Zhou, Yao Wan, Xiao Liu, Li Li, Hao Wu, Jin Liu, and Xin Jiang. 2021. Syncobert: Syntax-guided multi-modal contrastive pre-training for code representation. arXiv preprint arXiv: 2108.04556. tiano, and Geoffrey Irving. 2019. Fine-tuning language models from hum… find
  28. Xin Wang, Pingyi Zhou, Yasheng Wang, Xiao Liu, Jin Liu, and Hao Wu. 2021a. Servicebert: A pre-trained model for web service tagging and recommendation. In Service-Oriented Computing - 19th International Conference, ICSOC 2021, Virtual Event, November 22-25, 2021, Proceedings, volume 13121 of Lecture… find
  29. Yue Wang, Weishi Wang, Shafiq R. Joty, and Steven C. H. Hoi. 2021b. Codet5: Identifier-aware unified pre-trained encoder-decoder models for code understanding and generation. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, EMNLP 2021, Virtual Event / Punta… find
  30. Bolin Wei, Ge Li, Xin Xia, Zhiyi Fu, and Zhi Jin. 2019. Code generation as a dual task of code summarization. In Advances in Neural Information Processing Systems 32: Annual Conference on Neural Information Processing Systems 2019, NeurIPS 2019, December 8-14, 2019, Vancouver, BC, Canada, pages 6559… find
  31. Ronald J. Williams. 1992. Simple statistical gradient-following algorithms for connectionist reinforcement learning. Mach. Learn., 8:229–256. find
  32. Ziyu Yao, Jayavardhan Reddy Peddamail, and Huan Sun. 2019. Coacor: Code annotation for code retrieval with reinforcement learning. In The World Wide Web Conference, WWW 2019, San Francisco, CA, USA, May 13-17, 2019, pages 2203–2214. ACM. find
  33. Michihiro Yasunaga and Percy Liang. 2020. Graph-based, self-supervised program repair from diagnostic feedback. In Proceedings of the 37th International Conference on Machine Learning, ICML 2020, 13-18 July 2020, Virtual Event, volume 119 of Proceedings of Machine Learning Research, pages 10799–1080… find
  34. Michihiro Yasunaga and Percy Liang. 2021. Break-it-fix-it: Unsupervised learning for program repair. In Proceedings of the 38th International Conference on Machine Learning, ICML 2021, 18-24 July 2021, Virtual Event, volume 139 of Proceedings of Machine Learning Research, pages 11941–11952. PMLR. find
  35. Pengcheng Yin and Graham Neubig. 2017. A syntactic neural model for general-purpose code generation. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, ACL 2017, Vancouver, Canada, July 30 - August 4, Volume 1: Long Papers, pages 440–450. Association for Comp…
  36. Daniel M. Ziegler, Nisan Stiennon, Jeffrey Wu, Tom B. Brown, Alec Radford, Dario Amodei, Paul F. Chris- find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wang-2022-compilable-neural-code-generation-compiler/