Krony-PT: GPT2 compressed with Kronecker Products

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ayad, Mohamed Ayoub Ben, Mitrovic, Jelena, Granitzer, Michael
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914068713963520
author Ayad, Mohamed Ayoub Ben
Mitrovic, Jelena
Granitzer, Michael
author_facet Ayad, Mohamed Ayoub Ben
Mitrovic, Jelena
Granitzer, Michael
contents We introduce Krony-PT, a compression technique for GPT-2 based on Kronecker products. We specifically target the feed-forward weights of each transformer block, and systematically compress the feed-forward layer matrices to various degrees. We introduce a modified Van Loan decomposition to initialize new Kronecker factors, and also propose a new pruning-based initialization technique. Our method compresses the original 124M-parameter GPT-2 to various smaller models, ranging from 80M to 96M. Our 81M model variant outperforms DistilGPT2 on next-token prediction across all standard language modeling datasets, and shows competitive or comparable performance with significantly larger Kronecker-based compressions of GPT-2.
format Preprint
id arxiv_https___arxiv_org_abs_2412_12351
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Krony-PT: GPT2 compressed with Kronecker Products
Ayad, Mohamed Ayoub Ben
Mitrovic, Jelena
Granitzer, Michael
Machine Learning
Computation and Language
We introduce Krony-PT, a compression technique for GPT-2 based on Kronecker products. We specifically target the feed-forward weights of each transformer block, and systematically compress the feed-forward layer matrices to various degrees. We introduce a modified Van Loan decomposition to initialize new Kronecker factors, and also propose a new pruning-based initialization technique. Our method compresses the original 124M-parameter GPT-2 to various smaller models, ranging from 80M to 96M. Our 81M model variant outperforms DistilGPT2 on next-token prediction across all standard language modeling datasets, and shows competitive or comparable performance with significantly larger Kronecker-based compressions of GPT-2.
title Krony-PT: GPT2 compressed with Kronecker Products
topic Machine Learning
Computation and Language
url https://arxiv.org/abs/2412.12351