paper-with-me

Papers

PCEE-BERT: Accelerating BERT Inference via Patient and Confident Early Exiting

2022-01-16 · ACL ARR January 2022 1 · Anonymous

BERT and other pre-trained language models (PLMs) are ubiquitous in the modern NLP. Even though PLMs are the state-of-the-art (SOTA) models for almost every NLP task \citep{Qiu2020PretrainedMF}, the significant latency during inference forbids more widely industrial usage. In this work, we propose \underline{P}atient and \underline{C}onfident \underline{E}arly \underline{E}xiting BERT (PCEE-BERT), an off-the-shelf sample-dependent early exiting method that can work with different PLMs and can also work along with popular model compression methods. With a multi-exit BERT as the backbone model, PCEE-BERT will make the early exiting decision if enough numbers (patience parameter) of consecutive intermediate layers are confident about their predictions. The entropy value measures the confidence level of an intermediate layer's prediction. Experiments on the GLUE benchmark demonstrate that our method outperforms previous SOTA early exiting methods. Ablation studies show that: (a) our method performs consistently well on other PLMs, such as ALBERT and TinyBERT; (b) PCEE-BERT can make achieve different speed-up ratios by adjusting the patience parameter and the confidence threshold.

📄 PDF Abstract BibTeX

Code (0)

등록된 구현이 없습니다.

Tasks

Model Compression

Methods 이 논문이 사용한 방법론

Multi-Head Attention 설명 없음
Attention 설명 없음
Linear Layer A Linear Layer is a projection $\mathbf{XW + b}$.
LAMB LAMB is a a layerwise adaptive large batch optimization technique. It provides a strategy for adapting the learning rate in large batch settings. LAMB uses…
Dense Connections Dense Connections, or Fully Connected Connections, are a type of layer in a deep neural network that use a linear operation where every input is connected to every output…
Adam 설명 없음
Attention Dropout Attention Dropout is a type of dropout used in attention-based architectures, where elements are randomly dropped out of the…
Linear Warmup With Linear Decay Linear Warmup With Linear Decay is a learning rate schedule in which we increase the learning rate linearly for $n$ updates and then linearly decay afterwards.

Similar Papers 제목 키워드 기반

PCEE-BERT: Accelerating BERT Inference via Patient and Confident Early Exiting

2022-07-01 · Findings (NAACL) 2022 7 · Zhen Zhang, Wei Zhu, Jinfan Zhang, Peng Wang 외

BERT and other pretrained language models (PLMs) are ubiquitous in modern NLP. Even though PLMs are the state-of-the-art (SOTA) models for almost every NLP task (CITATION), the significant latency during inference prohib…

Model Compression

Elbert: Fast Albert with Confidence-Window Based Early Exit

2021-07-01 · Keli Xie, Siyuan Lu, Meiqi Wang, Zhongfeng Wang

Despite the great success in Natural Language Processing (NLP) area, large pre-trained language models like BERT are not well-suited for resource-constrained or real-time applications owing to the large number of paramet…

Decision Making

VEE-BERT: Accelerating BERT Inference for Named Entity Recognition via Vote Early Exiting

2022-01-16 · ACL ARR January 2022 1 · Anonymous

Named entity recognition (NER) is of great importance for a wide range of tasks, such as medical health record understanding, document analysis, dialogue understanding. BERT and its variants are the most performing model…

Dialogue Understandingnamed-entity-recognitionNamed Entity RecognitionNamed Entity Recognition (NER)+1

DeeBERT: Dynamic Early Exiting for Accelerating BERT Inference

2020-04-27 · ACL 2020 6 · Ji Xin, Raphael Tang, Jaejun Lee, Yao-Liang Yu 외

Large-scale pre-trained language models such as BERT have brought significant improvements to NLP applications. However, they are also notorious for being slow in inference, which makes them difficult to deploy in real-t…

PoWER-BERT: Accelerating BERT Inference via Progressive Word-vector Elimination

2020-01-24 · ICML 2020 1 · Saurabh Goyal, Anamitra R. Choudhury, Saurabh M. Raje, Venkatesan T. Chakaravarthy 외

We develop a novel method, called PoWER-BERT, for improving the inference time of the popular BERT model, while maintaining the accuracy. It works by: a) exploiting redundancy pertaining to word-vectors (intermediate enc…

General ClassificationNatural Language Understanding