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
Incorporating KenLM language model with HuggingFace implementation of Wav2Vec2CTC Model using beam search decoding

Wav2Vec2CTC With KenLM Using KenLM ARPA language model with beam search to decode audio files and show the most probable transcription. Assuming you'v

farisalasmary 65 Sep 21, 2022
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
Beyond Paragraphs: NLP for Long Sequences

Beyond Paragraphs: NLP for Long Sequences

AI2 338 Dec 02, 2022
PyTorch code for EMNLP 2019 paper "LXMERT: Learning Cross-Modality Encoder Representations from Transformers".

LXMERT: Learning Cross-Modality Encoder Representations from Transformers Our servers break again :(. I have updated the links so that they should wor

Hao Tan 838 Dec 19, 2022
A multi-lingual approach to AllenNLP CoReference Resolution along with a wrapper for spaCy.

Crosslingual Coreference Coreference is amazing but the data required for training a model is very scarce. In our case, the available training for non

Pandora Intelligence 71 Jan 04, 2023
DataCLUE: 国内首个以数据为中心的AI测评(含模型分析报告)

DataCLUE 以数据为中心的AI测评(DataCLUE) DataCLUE: A Chinese Data-centric Language Evaluation Benchmark 内容导引 章节 描述 简介 介绍以数据为中心的AI测评(DataCLUE)的背景 任务描述 任务描述 实验结果

CLUE benchmark 135 Dec 22, 2022
Random Directed Acyclic Graph Generator

DAG_Generator Random Directed Acyclic Graph Generator verison1.0 简介 工作流通常由DAG(有向无环图)来定义,其中每个计算任务$T_i$由一个顶点(node,task,vertex)表示。同时,任务之间的每个数据或控制依赖性由一条加权

Livion 17 Dec 27, 2022
GraphNLI: A Graph-based Natural Language Inference Model for Polarity Prediction in Online Debates

GraphNLI: A Graph-based Natural Language Inference Model for Polarity Prediction in Online Debates Vibhor Agarwal, Sagar Joglekar, Anthony P. Young an

Vibhor Agarwal 2 Jun 30, 2022
🌐 Translation microservice powered by AI

Dot Translate 🌐 A microservice for quick and local translation using A.I. This service starts a local webserver used for neural machine translation.

Dot HQ 48 Nov 22, 2022
تولید اسم های رندوم فینگیلیش

karafs کرفس تولید اسم های رندوم فینگیلیش installation ➜ pip install karafs usage دو زبانه ➜ karafs -n 10 توت فرنگی بی ناموس toot farangi-ye bi_namoos

Vaheed NÆINI (9E) 36 Nov 24, 2022
Under the hood working of transformers, fine-tuning GPT-3 models, DeBERTa, vision models, and the start of Metaverse, using a variety of NLP platforms: Hugging Face, OpenAI API, Trax, and AllenNLP

Transformers-for-NLP-2nd-Edition @copyright 2022, Packt Publishing, Denis Rothman Contact me for any question you have on LinkedIn Get the book on Ama

Denis Rothman 150 Dec 23, 2022
Kashgari is a production-level NLP Transfer learning framework built on top of tf.keras for text-labeling and text-classification, includes Word2Vec, BERT, and GPT2 Language Embedding.

Kashgari Overview | Performance | Installation | Documentation | Contributing 🎉 🎉 🎉 We released the 2.0.0 version with TF2 Support. 🎉 🎉 🎉 If you

Eliyar Eziz 2.3k Dec 29, 2022
Final Project Bootcamp Zero

The Quest (Pygame) Descripción Este es el repositorio de código The-Quest para el proyecto final Bootcamp Zero de KeepCoding. El juego consiste en la

Seven-z01 1 Mar 02, 2022
State of the Art Natural Language Processing

Spark NLP: State of the Art Natural Language Processing Spark NLP is a Natural Language Processing library built on top of Apache Spark ML. It provide

John Snow Labs 3k Jan 05, 2023
Weaviate demo with the text2vec-openai module

Weaviate demo with the text2vec-openai module This repository contains an example of how to use the Weaviate text2vec-openai module. When using this d

SeMI Technologies 11 Nov 11, 2022
Saptak Bhoumik 14 May 24, 2022
Search msDS-AllowedToActOnBehalfOfOtherIdentity

前言 现在进行RBCD的攻击手段主要是搜索mS-DS-CreatorSID,如果机器的创建者是我们可控的话,那就可以修改对应机器的msDS-AllowedToActOnBehalfOfOtherIdentity,利用工具SharpAllowedToAct-Modify 那我们索性也试试搜索所有计算机

Jumbo 26 Dec 05, 2022
Use fastai-v2 with HuggingFace's pretrained transformers

FastHugs Use fastai v2 with HuggingFace's pretrained transformers, see the notebooks below depending on your task: Text classification: fasthugs_seq_c

Morgan McGuire 111 Nov 16, 2022
Code for CodeT5: a new code-aware pre-trained encoder-decoder model.

CodeT5: Identifier-aware Unified Pre-trained Encoder-Decoder Models for Code Understanding and Generation This is the official PyTorch implementation

Salesforce 564 Jan 08, 2023
official ( API ) for the zAmericanEnglish app in [ Google play ] and [ App store ]

official ( API ) for the zAmericanEnglish app in [ Google play ] and [ App store ]

Plugin 3 Jan 12, 2022