Fine-Tuning Pretrained Language Models: Weight Initializations, Data Orders, and Early Stopping
Language Technologies Institute, School of Computer Science, Carnegie Mellon University（卡内基梅隆大学计算机科学学院语言技术研究所）；Allen Institute for Artificial Intelligence（艾伦人工智能研究所）；Paul G. Allen School of Computer Science and Engineering, University of Washington（华盛顿大学保罗·G·艾伦计算机科学与工程学院）
Jesse Dodge, Gabriel Ilharco, Roy Schwartz, Ali Farhadi, Hannaneh Hajishirzi, Noah Smith
Fine-tuning pretrained contextual word embedding models to supervised downstream tasks has become commonplace in natural language processing. This process, however, is often brittle: even with the same hyperparameter values, distinct random seeds can lead to substantially different results. To better understand this phenomenon, we experiment with four datasets from the GLUE benchmark, fine-tuning BERT hundreds of times on each while varying only the random seeds. We find substantial performance increases compared to previously reported results, and we quantify how the performance of the best-found model varies as a function of the number of fine-tuning trials. Further, we examine two factors influenced by the choice of random seed: weight initialization and training data order. We find that both contribute comparably to the variance of out-of-sample performance, and that some weight initializations perform well across all tasks explored. On small datasets, we observe that many fine-tuning trials diverge part of the way through training, and we offer best practices for practitioners to stop training less promising runs early. We publicly release all of our experimental data, including training and validation scores for 2,100 trials, to encourage further analysis of training dynamics during fine-tuning.
We present IndoBERTweet, the first large-scale pretrained model for Indonesian Twitter that is trained by extending a monolingually-trained Indonesian BERT model with additive domain-specific vocabulary. We focus in particular on efficient model adaptation under vocabulary mismatch, and benchmark different ways of initializing the BERT embedding layer for new word types. We find that initializing with the average BERT subword embedding makes pretraining five times faster, and is more effective than proposed methods for vocabulary adaptation in terms of extrinsic evaluation over seven Twitter-b…
We investigate the effectiveness of ensembles of pretrained transformer-based language models on short answer questions using the Kaggle Automated Short Answer Scoring dataset. We fine-tune a collection of popular small, base, and large pretrained transformer-based language models, and train one feature-base model on the dataset with the aim of testing ensembles of these models. We used an early stopping mechanism and hyperparameter optimization in training. We observe that generally that the larger models perform slightly better, however, they still fall short of state-of-the-art results one …
The popularity of pretrained language models in natural language processing systems calls for a careful evaluation of such models in down-stream tasks, which have a higher potential for societal impact. The evaluation of such systems usually focuses on accuracy measures. Our findings in this paper call for attention to be paid to fairness measures as well. Through the analysis of more than a dozen pretrained language models of varying sizes on two toxic text classification tasks (English), we demonstrate that focusing on accuracy measures alone can lead to models with wide variation in fairnes…
Large pretrained language models (LMs) have become the central building block of many NLP applications. Training these models requires ever more computational resources and most of the existing models are trained on English text only. It is exceedingly expensive to train these models in other languages. To alleviate this problem, we introduce a novel method -- called WECHSEL -- to efficiently and effectively transfer pretrained LMs to new languages. WECHSEL can be applied to any model which uses subword-based tokenization and learns an embedding for each subword. The tokenizer of the source mo…
Department of Computer Science and Technology, Tsinghua University（清华大学计算机科学与技术系）
In recent years, researchers tend to pre-train ever-larger language models to explore the upper limit of deep models. However, large language model pre-training costs intensive computational resources and most of the models are trained from scratch without reusing the existing pre-trained models, which is wasteful. In this paper, we propose bert2BERT, which can effectively transfer the knowledge of an existing smaller pre-trained model (e.g., BERT_BASE) to a large model (e.g., BERT_LARGE) through parameter initialization and significantly improve the pre-training efficiency of the large model.…
Product key memory (PKM) proposed by Lample et al. (2019) enables to improve prediction accuracy by increasing model capacity efficiently with insignificant computational overhead. However, their empirical application is only limited to causal language modeling. Motivated by the recent success of pretrained language models (PLMs), we investigate how to incorporate large PKM into PLMs that can be finetuned for a wide variety of downstream NLP tasks. We define a new memory usage metric, and careful observation using this metric reveals that most memory slots remain outdated during the training o…
PMSS: Pretrained Matrices Skeleton Selection for LLM Fine-tuning
Peking University（北京大学）；XiaoMi AI Lab（小米AI实验室）；Gaoling School of Artificial Intelligence, Renmin University of China（中国人民大学高瓴人工智能学院）；University of Electronic Science and Technology of China（电子科技大学）
A Comprehensive Survey on Pretrained Foundation Models: A History from BERT to ChatGPT
Michigan State University（密歇根州立大学）；Lehigh University（理海大学）；Beihang University（北京航空航天大学）；Macquarie University（麦考瑞大学）；Nanyang Technological University（南洋理工大学）；University of California San Diego（加利福尼亚大学圣地亚哥分校）；Salesforce AI Research（Salesforce人工智能研究院）；Duke University（杜克大学）；University of Illinois at Chicago（伊利诺伊大学芝加哥分校）
Word embeddings are the interface between the world of discrete units of text processing and the continuous, differentiable world of neural networks. In this work, we examine various random and pretrained initialization methods for embeddings used in deep networks and their effect on the performance on four NLP tasks with both recurrent and convolutional architectures. We confirm that pretrained embeddings are a little better than random initialization, especially considering the speed of learning. On the other hand, we do not see any significant difference between various methods of random in…
We propose a novel and simple method for semi-supervised text classification. The method stems from the hypothesis that a classifier with pretrained word embeddings always outperforms the same classifier with randomly initialized word embeddings, as empirically observed in NLP tasks. Our method first builds two sets of classifiers as a form of model ensemble, and then initializes their word embeddings differently: one using random, the other using pretrained word embeddings. We focus on different predictions between the two classifiers on unlabeled data while following the self-training framew…
Institute of Computer Science, Polish Academy of Sciences（波兰科学院计算机科学研究所）；AGH University of Science and Technology（AGH科技大学）
BERT-based models are currently used for solving nearly all Natural Language Processing (NLP) tasks and most often achieve state-of-the-art results. Therefore, the NLP community conducts extensive research on understanding these models, but above all on designing effective and efficient training procedures. Several ablation studies investigating how to train BERT-like models have been carried out, but the vast majority of them concerned only the English language. A training procedure designed for English does not have to be universal and applicable to other especially typologically different l…
Pretrained models are the standard starting point for training. This approach consistently outperforms the use of a random initialization. However, pretraining is a costly endeavour that few can undertake. In this paper, we create better base models at hardly any cost, by fusing multiple existing fine tuned models into one. Specifically, we fuse by averaging the weights of these models. We show that the fused model results surpass the pretrained model ones. We also show that fusing is often better than intertraining. We find that fusing is less dependent on the target task. Furthermore, weight…
Neural networks have excelled at many NLP tasks, but there remain open questions about the performance of pretrained distributed word representations and their interaction with weight initialization and other hyperparameters. We address these questions empirically using attention-based sequence-to-sequence models for natural language inference (NLI). Specifically, we compare three types of embeddings: random, pretrained (GloVe, word2vec), and retrofitted (pretrained plus WordNet information). We show that pretrained embeddings outperform both random and retrofitted ones in a large NLI corpus. …
Parameter Efficient Fine-tuning via Explained Variance Adaptation
ELLIS Unit, LIT AI Lab, Institute for Machine Learning, JKU Linz（ELLIS单位、LIT人工智能实验室、机器学习研究所、JKU林茨）；University College London（伦敦大学学院）；EMMI AI, Linz（EMMI AI、林茨）；NXAI GmbH, Linz, Austria（NXAI GmbH、林茨、奥地利）
本文提出 EVA 初始化方法，以激活方差最大方向构造 LoRA 子空间并支持自适应秩，在多类任务上实现更快收敛、更高平均得分及更少可训练参数。
Most Transformer language models are primarily pretrained on English text, limiting their use for other languages. As the model sizes grow, the performance gap between English and other languages with fewer compute and data resources increases even further. Consequently, more resource-efficient training methods are needed to bridge the gap for languages with fewer resources available. To address this problem, we introduce a cross-lingual and progressive transfer learning approach, called CLP-Transfer, that transfers models from a source language, for which pretrained models are publicly availa…
National Technical University of Athens（雅典国立技术大学）；Athena Research Center（雅典娜研究中心）；Behavioral Signal Technologies（行为信号技术公司）
In this work we explore Unsupervised Domain Adaptation (UDA) of pretrained language models for downstream tasks. We introduce UDALM, a fine-tuning procedure, using a mixed classification and Masked Language Model loss, that can adapt to the target domain distribution in a robust and sample efficient manner. Our experiments show that performance of models trained with the mixed loss scales with the amount of available target data and the mixed loss can be effectively used as a stopping criterion during UDA training. Furthermore, we discuss the relationship between A-distance and the target erro…
