Enhancing Deep Learning Transpilation with LLMs: Automation and Semantic Consistency
Deep learning (DL) transpilers perform source-to-source model conversion between DL frameworks, simplifying deployment. However, they often fail to bridge operator semantic inconsistencies, resulting in API incompatibilities (e.g., incompatible parameters, missing operators), which cause conversion failures or performance drops
Authors
- Zhijie Jiang (ORCID: https://orcid.org/0000-0002-7076-4239)
- Haifang Zhou (ORCID: https://orcid.org/0000-0001-7774-5623)
- Xingpei Li
- Zhouyang Jia (ORCID: https://orcid.org/0000-0002-2533-4547)
- Shanshan Li (ORCID: https://orcid.org/0000-0003-0798-974X)
- Yuanliang Zhang (ORCID: https://orcid.org/0009-0006-6042-7489)
- Si Zheng (ORCID: https://orcid.org/0009-0002-6363-859X)
Institutions
- National University of Defense Technology (CN)
Publication Details
- Published
- 2025-06-23
- DOI
- https://doi.org/10.1145/3696630.3731431
- Primary Topic
- Natural Language Processing Techniques
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Natural Science Foundation of China