On-the-fly knowledge distillation model for sentence embedding

In this dissertation, we run experimental study to investigate the performance of sentence embedding using an on-the-fly knowledge distillation model based on DistillCSE framework. This model utilizes SimCSE as the initial teacher model. After a certain number of training steps, it caches an interm...

Full description

Bibliographic Details
Main Author: Zhu, Xuchun
Other Authors: Lihui Chen
Format: Thesis-Master by Coursework
Language:English
Published: Nanyang Technological University 2024
Subjects:
Online Access:https://hdl.handle.net/10356/174236
_version_ 1826110780847685632
author Zhu, Xuchun
author2 Lihui Chen
author_facet Lihui Chen
Zhu, Xuchun
author_sort Zhu, Xuchun
collection NTU
description In this dissertation, we run experimental study to investigate the performance of sentence embedding using an on-the-fly knowledge distillation model based on DistillCSE framework. This model utilizes SimCSE as the initial teacher model. After a certain number of training steps, it caches an intermediate model and employs it as a new teacher model for knowledge distillation. This process is repeated several times to obtain the desired on-the-fly knowledge distilled student model. This model employs a novel approach to knowledge distillation, potentially offering advantages such as reducing training time and achieving performance close to the original teacher model. In some cases, after fine-tuning, it may even surpass the performance of the original teacher model for specific tasks.
first_indexed 2024-10-01T02:40:06Z
format Thesis-Master by Coursework
id ntu-10356/174236
institution Nanyang Technological University
language English
last_indexed 2024-10-01T02:40:06Z
publishDate 2024
publisher Nanyang Technological University
record_format dspace
spelling ntu-10356/1742362024-03-29T15:43:33Z On-the-fly knowledge distillation model for sentence embedding Zhu, Xuchun Lihui Chen School of Electrical and Electronic Engineering ELHCHEN@ntu.edu.sg Computer and Information Science On-the-fly model Knowledge distillation Sentence embeddings SimCSE DistillCSE In this dissertation, we run experimental study to investigate the performance of sentence embedding using an on-the-fly knowledge distillation model based on DistillCSE framework. This model utilizes SimCSE as the initial teacher model. After a certain number of training steps, it caches an intermediate model and employs it as a new teacher model for knowledge distillation. This process is repeated several times to obtain the desired on-the-fly knowledge distilled student model. This model employs a novel approach to knowledge distillation, potentially offering advantages such as reducing training time and achieving performance close to the original teacher model. In some cases, after fine-tuning, it may even surpass the performance of the original teacher model for specific tasks. Master's degree 2024-03-25T01:05:10Z 2024-03-25T01:05:10Z 2024 Thesis-Master by Coursework Zhu, X. (2024). On-the-fly knowledge distillation model for sentence embedding. Master's thesis, Nanyang Technological University, Singapore. https://hdl.handle.net/10356/174236 https://hdl.handle.net/10356/174236 en D-258-22231-05829 application/pdf Nanyang Technological University
spellingShingle Computer and Information Science
On-the-fly model
Knowledge distillation
Sentence embeddings
SimCSE
DistillCSE
Zhu, Xuchun
On-the-fly knowledge distillation model for sentence embedding
title On-the-fly knowledge distillation model for sentence embedding
title_full On-the-fly knowledge distillation model for sentence embedding
title_fullStr On-the-fly knowledge distillation model for sentence embedding
title_full_unstemmed On-the-fly knowledge distillation model for sentence embedding
title_short On-the-fly knowledge distillation model for sentence embedding
title_sort on the fly knowledge distillation model for sentence embedding
topic Computer and Information Science
On-the-fly model
Knowledge distillation
Sentence embeddings
SimCSE
DistillCSE
url https://hdl.handle.net/10356/174236
work_keys_str_mv AT zhuxuchun ontheflyknowledgedistillationmodelforsentenceembedding