Code2code Generation Using Language Models

Student: Kozlova Okhana

Supervisor:

Faculty: Faculty of Informatics, Mathematics, and Computer Science (HSE Nizhny Novgorod)

Educational Programme: Data Mining (Master)

Final Grade: 9

Year of Graduation: 2024

This study explores methods for solving the task of code-to-code generation using neural network architectures. Techniques for accelerating inference and compressing models to enhance their performance on CPUs and GPUs were examined and applied. СodeT5 and CodeT5+ neural network architectures were trained and tested on the CodeXGLUE dataset, utilizing speculative decoding and tensor decomposition methods, resulting in a 2.24 times speed increase with a CodeBLEU metric loss of 2.96.

Full text (added May 29, 2024)

Student Theses at HSE must be completed in accordance with the University Rules and regulations specified by each educational programme.

Summaries of all theses must be published and made freely available on the HSE website.

The full text of a thesis can be published in open access on the HSE website only if the authoring student (copyright holder) agrees, or, if the thesis was written by a team of students, if all the co-authors (copyright holders) agree. After a thesis is published on the HSE website, it obtains the status of an online publication.

Student theses are objects of copyright and their use is subject to limitations in accordance with the Russian Federation’s law on intellectual property.

In the event that a thesis is quoted or otherwise used, reference to the author’s name and the source of quotation is required.

Search all student theses