@inproceedings{marchisio-etal-2023-mini,
title = "Mini-Model Adaptation: Efficiently Extending Pretrained Models to New Languages via Aligned Shallow Training",
author = "Marchisio, Kelly and
Lewis, Patrick and
Chen, Yihong and
Artetxe, Mikel",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Findings of the Association for Computational Linguistics: ACL 2023",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-acl.338/",
doi = "10.18653/v1/2023.findings-acl.338",
pages = "5474--5490",
abstract = "Prior work shows that it is possible to expand pretrained Masked Language Models (MLMs) to new languages by learning a new set of embeddings, while keeping the transformer body frozen. Despite learning a small subset of parameters, this approach is not compute-efficient, as training the new embeddings requires a full forward and backward pass over the entire model. We propose mini-model adaptation, a compute-efficient alternative that builds a shallow mini-model from a fraction of a large model`s parameters. New language-specific embeddings can then be efficiently trained over the mini-model and plugged into the aligned large model for rapid cross-lingual transfer. We explore two approaches to learn mini-models: MINIJOINT, which jointly pretrains the primary model and the mini-model using a single transformer with a secondary MLM head at a middle layer; and MINIPOST, where we start from a regular pretrained model, build a mini-model by extracting and freezing a few layers, and learn a small number of parameters on top. Experiments on XNLI, MLQA and PAWS-X show that mini-model adaptation matches the performance of the standard approach using up to 2.3x less compute on average."
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="marchisio-etal-2023-mini">
<titleInfo>
<title>Mini-Model Adaptation: Efficiently Extending Pretrained Models to New Languages via Aligned Shallow Training</title>
</titleInfo>
<name type="personal">
<namePart type="given">Kelly</namePart>
<namePart type="family">Marchisio</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Patrick</namePart>
<namePart type="family">Lewis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yihong</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mikel</namePart>
<namePart type="family">Artetxe</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2023-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2023</title>
</titleInfo>
<name type="personal">
<namePart type="given">Anna</namePart>
<namePart type="family">Rogers</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jordan</namePart>
<namePart type="family">Boyd-Graber</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Naoaki</namePart>
<namePart type="family">Okazaki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Toronto, Canada</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Prior work shows that it is possible to expand pretrained Masked Language Models (MLMs) to new languages by learning a new set of embeddings, while keeping the transformer body frozen. Despite learning a small subset of parameters, this approach is not compute-efficient, as training the new embeddings requires a full forward and backward pass over the entire model. We propose mini-model adaptation, a compute-efficient alternative that builds a shallow mini-model from a fraction of a large model‘s parameters. New language-specific embeddings can then be efficiently trained over the mini-model and plugged into the aligned large model for rapid cross-lingual transfer. We explore two approaches to learn mini-models: MINIJOINT, which jointly pretrains the primary model and the mini-model using a single transformer with a secondary MLM head at a middle layer; and MINIPOST, where we start from a regular pretrained model, build a mini-model by extracting and freezing a few layers, and learn a small number of parameters on top. Experiments on XNLI, MLQA and PAWS-X show that mini-model adaptation matches the performance of the standard approach using up to 2.3x less compute on average.</abstract>
<identifier type="citekey">marchisio-etal-2023-mini</identifier>
<identifier type="doi">10.18653/v1/2023.findings-acl.338</identifier>
<location>
<url>https://aclanthology.org/2023.findings-acl.338/</url>
</location>
<part>
<date>2023-07</date>
<extent unit="page">
<start>5474</start>
<end>5490</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Mini-Model Adaptation: Efficiently Extending Pretrained Models to New Languages via Aligned Shallow Training
%A Marchisio, Kelly
%A Lewis, Patrick
%A Chen, Yihong
%A Artetxe, Mikel
%Y Rogers, Anna
%Y Boyd-Graber, Jordan
%Y Okazaki, Naoaki
%S Findings of the Association for Computational Linguistics: ACL 2023
%D 2023
%8 July
%I Association for Computational Linguistics
%C Toronto, Canada
%F marchisio-etal-2023-mini
%X Prior work shows that it is possible to expand pretrained Masked Language Models (MLMs) to new languages by learning a new set of embeddings, while keeping the transformer body frozen. Despite learning a small subset of parameters, this approach is not compute-efficient, as training the new embeddings requires a full forward and backward pass over the entire model. We propose mini-model adaptation, a compute-efficient alternative that builds a shallow mini-model from a fraction of a large model‘s parameters. New language-specific embeddings can then be efficiently trained over the mini-model and plugged into the aligned large model for rapid cross-lingual transfer. We explore two approaches to learn mini-models: MINIJOINT, which jointly pretrains the primary model and the mini-model using a single transformer with a secondary MLM head at a middle layer; and MINIPOST, where we start from a regular pretrained model, build a mini-model by extracting and freezing a few layers, and learn a small number of parameters on top. Experiments on XNLI, MLQA and PAWS-X show that mini-model adaptation matches the performance of the standard approach using up to 2.3x less compute on average.
%R 10.18653/v1/2023.findings-acl.338
%U https://aclanthology.org/2023.findings-acl.338/
%U https://doi.org/10.18653/v1/2023.findings-acl.338
%P 5474-5490
Markdown (Informal)
[Mini-Model Adaptation: Efficiently Extending Pretrained Models to New Languages via Aligned Shallow Training](https://aclanthology.org/2023.findings-acl.338/) (Marchisio et al., Findings 2023)
ACL