Guardado en:
Detalles Bibliográficos
Autores principales: Chang, Ernie, Lin, Pin-Jie, Li, Yang, Zhao, Changsheng, Kim, Daeil, Rabatin, Rastislav, Liu, Zechun, Shi, Yangyang, Chandra, Vikas
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:https://arxiv.org/abs/2409.14705
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866929509893144576
author Chang, Ernie
Lin, Pin-Jie
Li, Yang
Zhao, Changsheng
Kim, Daeil
Rabatin, Rastislav
Liu, Zechun
Shi, Yangyang
Chandra, Vikas
author_facet Chang, Ernie
Lin, Pin-Jie
Li, Yang
Zhao, Changsheng
Kim, Daeil
Rabatin, Rastislav
Liu, Zechun
Shi, Yangyang
Chandra, Vikas
contents Language model pretraining generally targets a broad range of use cases and incorporates data from diverse sources. However, there are instances where we desire a model that excels in specific areas without markedly compromising performance in other areas. A cost-effective and straightforward approach is sampling with low-dimensional data features, which allows to select large-scale pretraining data for domain-specific use cases. In this work, we revisit importance sampling with n-gram features consisting of multi-granular tokens, which strikes a good balance between sentence compression and representation capabilities. We observed the sampled data to have a high correlation with the target downstream task performance while preserving its effectiveness on other tasks. This leads to the proposed data sampling paradigm where language models can be pretrained more efficiently on selected documents. On eight benchmarks we demonstrate with $\sim$1% of the data, pretrained models perform on par with the full RefinedWeb data and outperform randomly selected samples for model sizes ranging from 125M to 1.5B.
format Preprint
id arxiv_https___arxiv_org_abs_2409_14705
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Target-Aware Language Modeling via Granular Data Sampling
Chang, Ernie
Lin, Pin-Jie
Li, Yang
Zhao, Changsheng
Kim, Daeil
Rabatin, Rastislav
Liu, Zechun
Shi, Yangyang
Chandra, Vikas
Computation and Language
Artificial Intelligence
Language model pretraining generally targets a broad range of use cases and incorporates data from diverse sources. However, there are instances where we desire a model that excels in specific areas without markedly compromising performance in other areas. A cost-effective and straightforward approach is sampling with low-dimensional data features, which allows to select large-scale pretraining data for domain-specific use cases. In this work, we revisit importance sampling with n-gram features consisting of multi-granular tokens, which strikes a good balance between sentence compression and representation capabilities. We observed the sampled data to have a high correlation with the target downstream task performance while preserving its effectiveness on other tasks. This leads to the proposed data sampling paradigm where language models can be pretrained more efficiently on selected documents. On eight benchmarks we demonstrate with $\sim$1% of the data, pretrained models perform on par with the full RefinedWeb data and outperform randomly selected samples for model sizes ranging from 125M to 1.5B.
title Target-Aware Language Modeling via Granular Data Sampling
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2409.14705