Scaling Stick-Breaking Attention: An Efficient Implementation and In-depth Study

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Tan, Shawn, Yang, Songlin, Courville, Aaron, Panda, Rameswar, Shen, Yikang
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912383304204288
author Tan, Shawn
Yang, Songlin
Courville, Aaron
Panda, Rameswar
Shen, Yikang
author_facet Tan, Shawn
Yang, Songlin
Courville, Aaron
Panda, Rameswar
Shen, Yikang
contents The self-attention mechanism traditionally relies on the softmax operator, necessitating positional embeddings like RoPE, or position biases to account for token order. But current methods using still face length generalisation challenges. We investigate an alternative attention mechanism based on the stick-breaking process in larger scale settings. The method works as follows: For each token before the current, we determine a break point, which represents the proportion of the stick, the weight of the attention, to allocate to the current token. We repeat this on the remaining stick, until all tokens are allocated a weight, resulting in a sequence of attention weights. This process naturally incorporates recency bias, which has linguistic motivations for grammar parsing. We study the implications of replacing the conventional softmax-based attention mechanism with stick-breaking attention. We then discuss implementation of numerically stable stick-breaking attention and adapt Flash Attention to accommodate this mechanism. When used as a drop-in replacement for current softmax+RoPE attention systems, we find that stick-breaking attention performs competitively with current methods on length generalisation and downstream tasks. Stick-breaking also performs well at length generalisation, allowing a model trained with $2^{11}$ context window to perform well at $2^{14}$ with perplexity improvements.
format Preprint
id arxiv_https___arxiv_org_abs_2410_17980
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Scaling Stick-Breaking Attention: An Efficient Implementation and In-depth Study
Tan, Shawn
Yang, Songlin
Courville, Aaron
Panda, Rameswar
Shen, Yikang
Machine Learning
Artificial Intelligence
Computation and Language
The self-attention mechanism traditionally relies on the softmax operator, necessitating positional embeddings like RoPE, or position biases to account for token order. But current methods using still face length generalisation challenges. We investigate an alternative attention mechanism based on the stick-breaking process in larger scale settings. The method works as follows: For each token before the current, we determine a break point, which represents the proportion of the stick, the weight of the attention, to allocate to the current token. We repeat this on the remaining stick, until all tokens are allocated a weight, resulting in a sequence of attention weights. This process naturally incorporates recency bias, which has linguistic motivations for grammar parsing. We study the implications of replacing the conventional softmax-based attention mechanism with stick-breaking attention. We then discuss implementation of numerically stable stick-breaking attention and adapt Flash Attention to accommodate this mechanism. When used as a drop-in replacement for current softmax+RoPE attention systems, we find that stick-breaking attention performs competitively with current methods on length generalisation and downstream tasks. Stick-breaking also performs well at length generalisation, allowing a model trained with $2^{11}$ context window to perform well at $2^{14}$ with perplexity improvements.
title Scaling Stick-Breaking Attention: An Efficient Implementation and In-depth Study
topic Machine Learning
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2410.17980