Institutional Repository of Vilnius University
Design and implementation of a parameter-efficient fine-tunning website for pre-defined ai model /
Abstract
dc:descriptionThis research addresses the software engineering challenges associated with parameter-efficient fine-tuning (PEFT) of large language models, particularly the complexity, cost and inefficiency of manual experimentation. The main contribution is the design and implementation of a web-based tool that helps researchers and students quickly evaluate PEFT methods and select appropriate hyperparameters, without performing computationally expensive training runs. By developing and combining predictive models and a rule-based recommendation system, the platform predicts training outcomes and guidelines the user for better outcome. As a result, developed system reduces computational overhead, making PEFT experimentation more accessible, while serving as an interactive learning tool to understand the impact of different PEFT approaches on model performance and efficiency.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Asiadouskaya, Hanna,
Subjects
dc:subject × 1Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- eng
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD277692702&prefLang=en_US