Understanding the Difficulty of Training Transformers

Overview

License PWC

Admin

Understanding the Difficulty of Training Transformers

Guided by our analyses, we propose Adaptive Model Initialization (Admin), which successfully stabilizes previously-diverged Transformer training and achieves better performance, without introducing additional hyper-parameters. Admin is adapted for better half-precision stability and can be reparameterized into the original Transformer.

We are in an early-release beta. Expect some adventures and rough edges.

Table of Contents

Introduction

What complicates Transformer training?

In our study, we go beyond gradient vanishing and identify an amplification effect that substantially influences Transformer training. Specifically, for each layer in a multi-layer Transformer, heavy dependency on its residual branch makes training unstable, yet light dependency leads to sub-optimal performance.

Dependency and Amplification Effect

Our analysis starts from the observation that Pre-LN is more robust than Post-LN, whereas Post-LN typically leads to a better performance. As shown in Figure 1, we find these two variants have different layer dependency patterns.

With further exploration, we find that for a N-layer residual network, after updating its parameters W to W*, its outputs change is proportion to the dependency on residual branches.

Intuitively, since a larger output change indicates a more unsmooth loss surface, the large dependency complicates training. Moreover, we propose Admin (adaptive model initialization), which starts the training from the area with a smoother surface. More details can be found in our paper.

Quick Start Guide

Our implementation is based on the fairseq package (python 3.6, torch 1.5/1.6 are recommended). It can be installed by:

git clone https://github.com/LiyuanLucasLiu/Transforemr-Clinic.git
cd fairseq
pip install --editable .

The guidance for reproducing our results is available at:

Specifically, our implementation requires to first set --init-type adaptive-profiling and use one GPU for this profiling stage, then set --init-type adaptive and start training.

Citation

Please cite the following papers if you found our model useful. Thanks!

Liyuan Liu, Xiaodong Liu, Jianfeng Gao, Weizhu Chen, and Jiawei Han (2020). Understanding the Difficulty of Training Transformers. Proc. 2020 Conf. on Empirical Methods in Natural Language Processing (EMNLP'20).

@inproceedings{liu2020admin,
  title={Understanding the Difficulty of Training Transformers},
  author = {Liu, Liyuan and Liu, Xiaodong and Gao, Jianfeng and Chen, Weizhu and Han, Jiawei},
  booktitle = {Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP 2020)},
  year={2020}
}

Xiaodong Liu, Kevin Duh, Liyuan Liu, and Jianfeng Gao (2020). Very Deep Transformers for Neural Machine Translation. arXiv preprint arXiv:2008.07772 (2020).

@inproceedings{liu_deep_2020,
 author = {Liu, Xiaodong and Duh, Kevin and Liu, Liyuan and Gao, Jianfeng},
 booktitle = {arXiv:2008.07772 [cs]},
 title = {Very Deep Transformers for Neural Machine Translation},
 year = {2020}
}
Owner
Liyuan Liu
Ph.D. Student @ DMG, UIUC
Liyuan Liu
Code for the Python code smells video on the ArjanCodes channel.

7 Python code smells This repository contains the code for the Python code smells video on the ArjanCodes channel (watch the video here). The example

55 Dec 29, 2022
Fully featured implementation of Routing Transformer

Routing Transformer A fully featured implementation of Routing Transformer. The paper proposes using k-means to route similar queries / keys into the

Phil Wang 246 Jan 02, 2023
A python package to fine-tune transformer-based models for named entity recognition (NER).

nerblackbox A python package to fine-tune transformer-based language models for named entity recognition (NER). Resources Source Code: https://github.

Felix Stollenwerk 13 Jul 30, 2022
A python gui program to generate reddit text to speech videos from the id of any post.

Reddit text to speech generator A python gui program to generate reddit text to speech videos from the id of any post. Current functionality Generate

Aadvik 17 Dec 19, 2022
Knowledge Oriented Programming Language

KoPL: 面向知识的推理问答编程语言 安装 | 快速开始 | 文档 KoPL全称 Knowledge oriented Programing Language, 是一个为复杂推理问答而设计的编程语言。我们可以将自然语言问题表示为由基本函数组合而成的KoPL程序,程序运行的结果就是问题的答案。目前,

THU-KEG 62 Dec 12, 2022
Library for fast text representation and classification.

fastText fastText is a library for efficient learning of word representations and sentence classification. Table of contents Resources Models Suppleme

Facebook Research 24.1k Jan 05, 2023
The code for the Subformer, from the EMNLP 2021 Findings paper: "Subformer: Exploring Weight Sharing for Parameter Efficiency in Generative Transformers", by Machel Reid, Edison Marrese-Taylor, and Yutaka Matsuo

Subformer This repository contains the code for the Subformer. To help overcome this we propose the Subformer, allowing us to retain performance while

Machel Reid 10 Dec 27, 2022
Creating a python chatbot that Starbucks users can text to place an order + help cut wait time of a normal coffee.

Creating a python chatbot that Starbucks users can text to place an order + help cut wait time of a normal coffee.

2 Jan 20, 2022
A Telegram bot to add notes to Flomo.

flomo bot 使用 Telegram 机器人发送笔记到你的 Flomo. 你需要有一台可访问 Telegram 的服务器。 Steps @BotFather 新建机器人,获取 token Flomo 官网获取 API,链接 https://flomoapp.com/mine?source=in

Zhen 44 Dec 30, 2022
The simple project to separate mixed voice (2 clean voices) to 2 separate voices.

Speech Separation The simple project to separate mixed voice (2 clean voices) to 2 separate voices. Result Example (Clisk to hear the voices): mix ||

vuthede 31 Oct 30, 2022
A Pytorch implementation of "Splitter: Learning Node Representations that Capture Multiple Social Contexts" (WWW 2019).

Splitter ⠀⠀ A PyTorch implementation of Splitter: Learning Node Representations that Capture Multiple Social Contexts (WWW 2019). Abstract Recent inte

Benedek Rozemberczki 201 Nov 09, 2022
문장단위로 분절된 나무위키 데이터셋. Releases에서 다운로드 받거나, tfds-korean을 통해 다운로드 받으세요.

Namuwiki corpus 문장단위로 미리 분절된 나무위키 코퍼스. 목적이 LM등에서 사용하기 위한 데이터셋이라, 링크/이미지/테이블 등등이 잘려있습니다. 문장 단위 분절은 kss를 활용하였습니다. 라이선스는 나무위키에 명시된 바와 같이 CC BY-NC-SA 2.0

Jeong Ukjae 16 Apr 02, 2022
Speach Recognitions

easy_meeting Добро пожаловать в интерфейс сервиса автопротоколирования совещаний Easy Meeting. Website - http://cf5c-62-192-251-83.ngrok.io/ Принципиа

Maksim 3 Feb 18, 2022
Text Normalization(文本正则化)

Text Normalization(文本正则化) 任务描述:通过机器学习算法将英文文本的“手写”形式转换成“口语“形式,例如“6ft”转换成“six feet”等 实验结果 XGBoost + bag-of-words: 0.99159 XGBoost+Weights+rules:0.99002

Jason_Zhang 0 Feb 26, 2022
Transformer related optimization, including BERT, GPT

This repository provides a script and recipe to run the highly optimized transformer-based encoder and decoder component, and it is tested and maintained by NVIDIA.

NVIDIA Corporation 1.7k Jan 04, 2023
Training code for Korean multi-class sentiment analysis

KoSentimentAnalysis Bert implementation for the Korean multi-class sentiment analysis 왜 한국어 감정 다중분류 모델은 거의 없는 것일까?에서 시작된 프로젝트 Environment: Pytorch, Da

Donghoon Shin 3 Dec 02, 2022
Enterprise Scale NLP with Hugging Face & SageMaker Workshop series

Workshop: Enterprise-Scale NLP with Hugging Face & Amazon SageMaker Earlier this year we announced a strategic collaboration with Amazon to make it ea

Philipp Schmid 161 Dec 16, 2022
Simple Text-To-Speech Bot For Discord

Simple Text-To-Speech Bot For Discord This is a very simple TTS bot for discord made with python. For this bot you need FFMPEG, see installation to se

1 Sep 26, 2022
BookNLP, a natural language processing pipeline for books

BookNLP BookNLP is a natural language processing pipeline that scales to books and other long documents (in English), including: Part-of-speech taggin

654 Jan 02, 2023
Use the power of GPT3 to execute any function inside your programs just by giving some doctests

gptrun Don't feel like coding today? Use the power of GPT3 to execute any function inside your programs just by giving some doctests. How is this diff

Roberto Abdelkader Martínez Pérez 11 Nov 11, 2022