Parameters vs FLOPs: Scaling Laws for Optimal Sparsity for Mixture-of-Experts Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Abnar, Samira, Shah, Harshay, Busbridge, Dan, Ali, Alaaeldin Mohamed Elnouby, Susskind, Josh, Thilak, Vimal
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912461458767872
author Abnar, Samira
Shah, Harshay
Busbridge, Dan
Ali, Alaaeldin Mohamed Elnouby
Susskind, Josh
Thilak, Vimal
author_facet Abnar, Samira
Shah, Harshay
Busbridge, Dan
Ali, Alaaeldin Mohamed Elnouby
Susskind, Josh
Thilak, Vimal
contents Scaling the capacity of language models has consistently proven to be a reliable approach for improving performance and unlocking new capabilities. Capacity can be primarily defined by two dimensions: the number of model parameters and the compute per example. While scaling typically involves increasing both, the precise interplay between these factors and their combined contribution to overall capacity remains not fully understood. We explore this relationship in the context of sparse Mixture-of-Experts (MoEs), which allow scaling the number of parameters without proportionally increasing the FLOPs per example. We investigate how varying the sparsity level, i.e., the fraction of inactive parameters, impacts model's performance during pretraining and downstream few-shot evaluation. We find that under different constraints (e.g., parameter size and total training compute), there is an optimal level of sparsity that improves both training efficiency and model performance. These results provide a better understanding of the impact of sparsity in scaling laws for MoEs and complement existing works in this area, offering insights for designing more efficient architectures.
format Preprint
id arxiv_https___arxiv_org_abs_2501_12370
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Parameters vs FLOPs: Scaling Laws for Optimal Sparsity for Mixture-of-Experts Language Models
Abnar, Samira
Shah, Harshay
Busbridge, Dan
Ali, Alaaeldin Mohamed Elnouby
Susskind, Josh
Thilak, Vimal
Machine Learning
Artificial Intelligence
Scaling the capacity of language models has consistently proven to be a reliable approach for improving performance and unlocking new capabilities. Capacity can be primarily defined by two dimensions: the number of model parameters and the compute per example. While scaling typically involves increasing both, the precise interplay between these factors and their combined contribution to overall capacity remains not fully understood. We explore this relationship in the context of sparse Mixture-of-Experts (MoEs), which allow scaling the number of parameters without proportionally increasing the FLOPs per example. We investigate how varying the sparsity level, i.e., the fraction of inactive parameters, impacts model's performance during pretraining and downstream few-shot evaluation. We find that under different constraints (e.g., parameter size and total training compute), there is an optimal level of sparsity that improves both training efficiency and model performance. These results provide a better understanding of the impact of sparsity in scaling laws for MoEs and complement existing works in this area, offering insights for designing more efficient architectures.
title Parameters vs FLOPs: Scaling Laws for Optimal Sparsity for Mixture-of-Experts Language Models
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2501.12370