Saved in:
Bibliographic Details
Main Authors: Kosson, Atli, Messmer, Bettina, Jaggi, Martin
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2410.23922
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912098270838784
author Kosson, Atli
Messmer, Bettina
Jaggi, Martin
author_facet Kosson, Atli
Messmer, Bettina
Jaggi, Martin
contents Learning Rate Warmup is a popular heuristic for training neural networks, especially at larger batch sizes, despite limited understanding of its benefits. Warmup decreases the update size $Δ\mathbf{w}_t = η_t \mathbf{u}_t$ early in training by using lower values for the learning rate $η_t$. In this work we argue that warmup benefits training by keeping the overall size of $Δ\mathbf{w}_t$ limited, counteracting large initial values of $\mathbf{u}_t$. Focusing on small-scale GPT training with AdamW/Lion, we explore the following question: Why and by which criteria are early updates $\mathbf{u}_t$ too large? We analyze different metrics for the update size including the $\ell_2$-norm, resulting directional change, and impact on the representations of the network, providing a new perspective on warmup. In particular, we find that warmup helps counteract large angular updates as well as a limited critical batch size early in training. Finally, we show that the need for warmup can be significantly reduced or eliminated by modifying the optimizer to explicitly normalize $\mathbf{u}_t$ based on the aforementioned metrics.
format Preprint
id arxiv_https___arxiv_org_abs_2410_23922
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Analyzing & Reducing the Need for Learning Rate Warmup in GPT Training
Kosson, Atli
Messmer, Bettina
Jaggi, Martin
Machine Learning
Learning Rate Warmup is a popular heuristic for training neural networks, especially at larger batch sizes, despite limited understanding of its benefits. Warmup decreases the update size $Δ\mathbf{w}_t = η_t \mathbf{u}_t$ early in training by using lower values for the learning rate $η_t$. In this work we argue that warmup benefits training by keeping the overall size of $Δ\mathbf{w}_t$ limited, counteracting large initial values of $\mathbf{u}_t$. Focusing on small-scale GPT training with AdamW/Lion, we explore the following question: Why and by which criteria are early updates $\mathbf{u}_t$ too large? We analyze different metrics for the update size including the $\ell_2$-norm, resulting directional change, and impact on the representations of the network, providing a new perspective on warmup. In particular, we find that warmup helps counteract large angular updates as well as a limited critical batch size early in training. Finally, we show that the need for warmup can be significantly reduced or eliminated by modifying the optimizer to explicitly normalize $\mathbf{u}_t$ based on the aforementioned metrics.
title Analyzing & Reducing the Need for Learning Rate Warmup in GPT Training
topic Machine Learning
url https://arxiv.org/abs/2410.23922