Investigating the Synergistic Effects of Dropout and Residual Connections on Language Model Training

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Qingyang, Ke, Weimao
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913526675668992
author Li, Qingyang
Ke, Weimao
author_facet Li, Qingyang
Ke, Weimao
contents This paper examines the pivotal role of dropout techniques in mitigating overfitting in language model training. It conducts a comprehensive investigation into the influence of variable dropout rates on both individual layers and residual connections within the context of language modeling. Our study conducts training of a decoder implementation on the classic Tiny Shakespeare data to examine the effects of the adjustments on training efficiency and validation error. Results not only confirm the benefits of dropout for regularization and residuals for convergence, but also reveal their interesting interactions. There exists an important trade-off between the depth of residual connections and the dropout on these connections for optimal deep neural network convergence and generalization.
format Preprint
id arxiv_https___arxiv_org_abs_2410_01019
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Investigating the Synergistic Effects of Dropout and Residual Connections on Language Model Training
Li, Qingyang
Ke, Weimao
Computation and Language
Machine Learning
This paper examines the pivotal role of dropout techniques in mitigating overfitting in language model training. It conducts a comprehensive investigation into the influence of variable dropout rates on both individual layers and residual connections within the context of language modeling. Our study conducts training of a decoder implementation on the classic Tiny Shakespeare data to examine the effects of the adjustments on training efficiency and validation error. Results not only confirm the benefits of dropout for regularization and residuals for convergence, but also reveal their interesting interactions. There exists an important trade-off between the depth of residual connections and the dropout on these connections for optimal deep neural network convergence and generalization.
title Investigating the Synergistic Effects of Dropout and Residual Connections on Language Model Training
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2410.01019