使用pytorch+transformers复现了SimCSE论文中的有监督训练和无监督训练方法

Related tags

Text Data & NLPSimCSE
Overview

SimCSE复现

项目描述

SimCSE是一种简单但是很巧妙的NLP对比学习方法,创新性地引入Dropout的方式,对样本添加噪声,从而达到对正样本增强的目的。 该框架的训练目的为:对于batch中的每个样本,拉近其与正样本之间的距离,拉远其与负样本之间的距离,使得模型能够在大规模无监督语料(也可以使用有监督的语料)中学习到文本相似关系。 详见论文:Simple Contrastive Learning of Sentence EmbeddingsSimCSE官方代码仓库

本项目使用pytorch+transformers复现了SimCSE论文中的有监督训练和无监督训练方法,并且在STS-B数据集上进行消融实验,评价指标为Spearman相关系数,预训练模型为Bert-base-uncased, 验证了SimCSE的有效性。在STS-B数据集上,有监督训练和无监督训练的复现效果如下表。

在无监督训练中,dropout=0.1时,复现效果比原文略差,但也比较接近。当dropout=0.2时,复现效果比原文略高。 ** 但在有监督训练中,不知是否由于batch size过小(原论文使用512),复现效果与论文的效果相差较远,后续会进行排查。 **

训练方法 learning rate batch size dropout Spearman’s correlation
原论文 无监督 3e-5 64 0.1 0.763
复现 无监督 3e-5 64 0.2 0.771
复现 无监督 3e-5 64 0.1 0.748
原论文 有监督 5e-5 512 0.1 0.816
复现 有监督 5e-5 64 0.1 0.764

运行环境

python==3.6、transformers==3.1.0、torch==1.6.0

项目结构

  • data:存放训练数据
    • stsbenchmark:STS-B数据集
      • sts-dev.csv:STS-B验证集
      • sts-test.csv:STS-B验测试集
    • nli_for_simcse.csv:数量275601为的NLI数据集
    • wiki1m_for_simcse.txt:维基百科上获取的100w的文本
  • output:输出目录
  • pretrain_model:预训练模型存放位置
  • script:脚本存放位置。
  • dataset.py
  • model.py:模型代码,包含有监督和无监督损失函数的计算方式
  • train.py:训练代码

使用方法

Quick Start

下载训练数据:

bash script/download_nli.sh
bash script/download_wiki.sh

无监督训练,运行脚本

bash script/run_unsup_train.sh

有监督训练,运行脚本

bash script/run_sup_train.sh

实验

无监督训练

从前四条实验数据中可以看到,较大的batch size在一定程度上可以增加模型的泛化性。

dropout为0.2的时候,训练效果比0.1与0.3更好,有可能dropout=0.1加入的噪声过小,而dropout=0.3加入的噪声过大,增强得到的样本与原始样本差异较大。

learning rate batch size dropout 在哪一步得到best checkpoint 验证集上的得分 测试集上的得分
3e-5 256 0.1 6000 0.800 0.761
3e-5 128 0.1 4200 0.799 0.747
3e-5 64 0.1 10900 0.803 0.748
3e-5 32 0.1 21300 0.787 0.714
3e-5 64 0.2 11200 0.811 0.771
3e-5 64 0.3 6300 0.781 0.745
1e-5 64 0.1 16400 0.798 0.751

有监督训练

有监督实验的复现结果未达到预期,超参数相同时,在验证集上的得分略高于无监督,但是在测试集上,得分基本没有差异。增大有监督训练的学习率,有监督的训练的得分略高于无监督训练, 但还是与论文声称的0.816相差较远,原论文使用512的batch size, 不知是否由于batch size的设置有关,后续会对有监督的训练代码进一步排查。

不过从训练曲线可以看到,有监督训练的收敛速度明显快于无监督训练,这也符合我们的认知。

训练方法 learning rate batch size dropout 在哪一步得到best checkpoint 验证集上的得分 测试集上的得分
无监督 3e-5 64 0.1 10900 0.803 0.748
有监督 3e-5 64 0.1 200 0.810 0.748
有监督 5e-5 64 0.1 2300 0.809 0.764
有监督 3e-5 32 0.1 200 0.808 0.743
有监督 5e-5 32 0.1 200 0.806 0.746

无监督训练过程中,验证集得分的变化曲线: avatar

有监督训练过程中,验证集得分的变化曲线: avatar

REFERENCE

TODO

  • 排查有监督学习的效果不符合预期的原因
Big Bird: Transformers for Longer Sequences

BigBird, is a sparse-attention based transformer which extends Transformer based models, such as BERT to much longer sequences. Moreover, BigBird comes along with a theoretical understanding of the c

Google Research 457 Dec 23, 2022
Tracking Progress in Natural Language Processing

Repository to track the progress in Natural Language Processing (NLP), including the datasets and the current state-of-the-art for the most common NLP tasks.

Sebastian Ruder 21.2k Dec 30, 2022
Generate product descriptions, blogs, ads and more using GPT architecture with a single request to TextCortex API a.k.a Hemingwai

TextCortex - HemingwAI Generate product descriptions, blogs, ads and more using GPT architecture with a single request to TextCortex API a.k.a Hemingw

TextCortex AI 27 Nov 28, 2022
Tutorial to pretrain & fine-tune a 🤗 Flax T5 model on a TPUv3-8 with GCP

Pretrain and Fine-tune a T5 model with Flax on GCP This tutorial details how pretrain and fine-tune a FlaxT5 model from HuggingFace using a TPU VM ava

Gabriele Sarti 41 Nov 18, 2022
Chinese NewsTitle Generation Project by GPT2.带有超级详细注释的中文GPT2新闻标题生成项目。

GPT2-NewsTitle 带有超详细注释的GPT2新闻标题生成项目 UpDate 01.02.2021 从网上收集数据,将清华新闻数据、搜狗新闻数据等新闻数据集,以及开源的一些摘要数据进行整理清洗,构建一个较完善的中文摘要数据集。 数据集清洗时,仅进行了简单地规则清洗。

logCong 785 Dec 29, 2022
Hostapd-mac-tod-acl - Setup a hostapd AP with MAC ToD ACL

A brief explanation This script provides a quick way to setup a Time-of-day (Tod

2 Feb 03, 2022
MEDIALpy: MEDIcal Abbreviations Lookup in Python

A small python package that allows the user to look up common medical abbreviations.

Aberystwyth Systems Biology 7 Nov 09, 2022
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
Binary LSTM model for text classification

Text Classification The purpose of this repository is to create a neural network model of NLP with deep learning for binary classification of texts re

Nikita Elenberger 1 Mar 11, 2022
🏖 Easy training and deployment of seq2seq models.

Headliner Headliner is a sequence modeling library that eases the training and in particular, the deployment of custom sequence models for both resear

Axel Springer Ideas Engineering GmbH 231 Nov 18, 2022
:P Some basic stuff I'm gonna use for my upcoming Agile Software Development and Devops

reverse-image-search-py bash script.sh img_name.jpg Requirements pip install requests pip install pyshorteners Dry run [ Sudhanva M 3 Dec 18, 2021

Bu Chatbot, Konya Bilim Merkezi Yen için tasarlanmış olan bir projedir.

chatbot Bu Chatbot, Konya Bilim Merkezi Yeni Ufuklar Sergisi için 2021 Yılında tasarlanmış olan bir projedir. Chatbot Python ortamında yazılmıştır. Sö

Emre Özkul 1 Feb 23, 2022
What are the best Systems? New Perspectives on NLP Benchmarking

What are the best Systems? New Perspectives on NLP Benchmarking In Machine Learning, a benchmark refers to an ensemble of datasets associated with one

Pierre Colombo 12 Nov 03, 2022
Arabic speech recognition, classification and text-to-speech.

klaam Arabic speech recognition, classification and text-to-speech using many advanced models like wave2vec and fastspeech2. This repository allows tr

ARBML 177 Dec 27, 2022
Synthetic data for the people.

zpy: Synthetic data in Blender. Website • Install • Docs • Examples • CLI • Contribute • Licence Abstract Collecting, labeling, and cleaning data for

Zumo Labs 253 Dec 21, 2022
A raytrace framework using taichi language

ti-raytrace The code use Taichi programming language Current implement acceleration lvbh disney brdf How to run First config your anaconda workspace,

蕉太狼 73 Dec 11, 2022
This repo contains simple to use, pretrained/training-less models for speaker diarization.

PyDiar This repo contains simple to use, pretrained/training-less models for speaker diarization. Supported Models Binary Key Speaker Modeling Based o

12 Jan 20, 2022
A simple implementation of N-gram language model.

About A simple implementation of N-gram language model. Requirements numpy Data preparation Corpus Training data for the N-gram model, a text file lik

4 Nov 24, 2021
Named-entity recognition using neural networks. Easy-to-use and state-of-the-art results.

NeuroNER NeuroNER is a program that performs named-entity recognition (NER). Website: neuroner.com. This page gives step-by-step instructions to insta

Franck Dernoncourt 1.6k Dec 27, 2022
CrossNER: Evaluating Cross-Domain Named Entity Recognition (AAAI-2021)

CrossNER is a fully-labeled collected of named entity recognition (NER) data spanning over five diverse domains (Politics, Natural Science, Music, Literature, and Artificial Intelligence) with specia

Zihan Liu 89 Nov 10, 2022