LLM Vocabulary Compression for Low-Compute Environments

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Vennam, Sreeram, Joishy, Anish, Kumaraguru, Ponnurangam
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909383137427456
author Vennam, Sreeram
Joishy, Anish
Kumaraguru, Ponnurangam
author_facet Vennam, Sreeram
Joishy, Anish
Kumaraguru, Ponnurangam
contents We present a method to compress the final linear layer of language models, reducing memory usage by up to 3.4x without significant performance loss. By grouping tokens based on Byte Pair Encoding (BPE) merges, we prevent materialization of the memory-intensive logits tensor. Evaluations on the TinyStories dataset show that our method performs on par with GPT-Neo and GPT2 while significantly improving throughput by up to 3x, making it suitable for low-compute environments.
format Preprint
id arxiv_https___arxiv_org_abs_2411_06371
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle LLM Vocabulary Compression for Low-Compute Environments
Vennam, Sreeram
Joishy, Anish
Kumaraguru, Ponnurangam
Computation and Language
Machine Learning
I.2.6; I.2.7
We present a method to compress the final linear layer of language models, reducing memory usage by up to 3.4x without significant performance loss. By grouping tokens based on Byte Pair Encoding (BPE) merges, we prevent materialization of the memory-intensive logits tensor. Evaluations on the TinyStories dataset show that our method performs on par with GPT-Neo and GPT2 while significantly improving throughput by up to 3x, making it suitable for low-compute environments.
title LLM Vocabulary Compression for Low-Compute Environments
topic Computation and Language
Machine Learning
I.2.6; I.2.7
url https://arxiv.org/abs/2411.06371