MultiTok: Variable-Length Tokenization for Efficient LLMs Adapted from LZW Compression

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Elias, Noel, Esfahanizadeh, Homa, Kale, Kaan, Vishwanath, Sriram, Medard, Muriel
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911619061121024
author Elias, Noel
Esfahanizadeh, Homa
Kale, Kaan
Vishwanath, Sriram
Medard, Muriel
author_facet Elias, Noel
Esfahanizadeh, Homa
Kale, Kaan
Vishwanath, Sriram
Medard, Muriel
contents Large language models have drastically changed the prospects of AI by introducing technologies for more complex natural language processing. However, current methodologies to train such LLMs require extensive resources including but not limited to large amounts of data, expensive machinery, and lengthy training. To solve this problem, this paper proposes a new tokenization method inspired by universal Lempel-Ziv-Welch data compression that compresses repetitive phrases into multi-word tokens. With MultiTok as a new tokenizing tool, we show that language models are able to be trained notably more efficiently while offering a similar accuracy on more succinct and compressed training data. In fact, our results demonstrate that MultiTok achieves a comparable performance to the BERT and GPT standards as both a stand-alone tokenizer and an add-on to existing tokenizers while also providing close to 2.5x faster training with more than 30% less training data.
format Preprint
id arxiv_https___arxiv_org_abs_2410_21548
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle MultiTok: Variable-Length Tokenization for Efficient LLMs Adapted from LZW Compression
Elias, Noel
Esfahanizadeh, Homa
Kale, Kaan
Vishwanath, Sriram
Medard, Muriel
Computation and Language
Information Theory
Machine Learning
Large language models have drastically changed the prospects of AI by introducing technologies for more complex natural language processing. However, current methodologies to train such LLMs require extensive resources including but not limited to large amounts of data, expensive machinery, and lengthy training. To solve this problem, this paper proposes a new tokenization method inspired by universal Lempel-Ziv-Welch data compression that compresses repetitive phrases into multi-word tokens. With MultiTok as a new tokenizing tool, we show that language models are able to be trained notably more efficiently while offering a similar accuracy on more succinct and compressed training data. In fact, our results demonstrate that MultiTok achieves a comparable performance to the BERT and GPT standards as both a stand-alone tokenizer and an add-on to existing tokenizers while also providing close to 2.5x faster training with more than 30% less training data.
title MultiTok: Variable-Length Tokenization for Efficient LLMs Adapted from LZW Compression
topic Computation and Language
Information Theory
Machine Learning
url https://arxiv.org/abs/2410.21548