ترغب بنشر مسار تعليمي؟ اضغط هنا

MLPruning: A Multilevel Structured Pruning Framework for Transformer-based Models

178   0   0.0 ( 0 )
 نشر من قبل Zhewei Yao
 تاريخ النشر 2021
  مجال البحث الهندسة المعلوماتية
والبحث باللغة English




اسأل ChatGPT حول البحث

Pruning is an effective method to reduce the memory footprint and computational cost associated with large natural language processing models. However, current approaches either only explore head pruning, which has a limited pruning ratio, or only focus on unstructured pruning, which has negligible effects on the real inference time and/or power consumption. To address these challenges, we develop a novel MultiLevel structured Pruning (MLPruning) framework, which uses three different levels of structured pruning: head pruning, row pruning, and block-wise sparse pruning. We propose using a learnable Top-k threshold, which employs an adaptive regularization to adjust the regularization magnitude adaptively, to select appropriate pruning ratios for different weight matrices. We also propose a two-step pipeline to combine block-wise pruning with head/row pruning to achieve high structured pruning ratios with minimum accuracy degradation. Our empirical results show that for bertbase, with textapprox20% of remaining weights, OURS can achieve an accuracy that is comparable to the full model on QQP/MNLI/squad, with up to textapprox3.69x speedup. Our framework has been open sourced~cite{codebase}.



قيم البحث

اقرأ أيضاً

Pre-trained large-scale language models have increasingly demonstrated high accuracy on many natural language processing (NLP) tasks. However, the limited weight storage and computational speed on hardware platforms have impeded the popularity of pre -trained models, especially in the era of edge computing. In this work, we propose an efficient transformer-based large-scale language representation using hardware-friendly block structure pruning. We incorporate the reweighted group Lasso into block-structured pruning for optimization. Besides the significantly reduced weight storage and computation, the proposed approach achieves high compression rates. Experimental results on different models (BERT, RoBERTa, and DistilBERT) on the General Language Understanding Evaluation (GLUE) benchmark tasks show that we achieve up to 5.0x with zero or minor accuracy degradation on certain task(s). Our proposed method is also orthogonal to existing compact pre-trained language models such as DistilBERT using knowledge distillation, since a further 1.79x average compression rate can be achieved on top of DistilBERT with zero or minor accuracy degradation. It is suitable to deploy the final compressed model on resource-constrained edge devices.
Text normalization (TN) and inverse text normalization (ITN) are essential preprocessing and postprocessing steps for text-to-speech synthesis and automatic speech recognition, respectively. Many methods have been proposed for either TN or ITN, rangi ng from weighted finite-state transducers to neural networks. Despite their impressive performance, these methods aim to tackle only one of the two tasks but not both. As a result, in a complete spoken dialog system, two separate models for TN and ITN need to be built. This heterogeneity increases the technical complexity of the system, which in turn increases the cost of maintenance in a production setting. Motivated by this observation, we propose a unified framework for building a single neural duplex system that can simultaneously handle TN and ITN. Combined with a simple but effective data augmentation method, our systems achieve state-of-the-art results on the Google TN dataset for English and Russian. They can also reach over 95% sentence-level accuracy on an internal English TN dataset without any additional fine-tuning. In addition, we also create a cleaned dataset from the Spoken Wikipedia Corpora for German and report the performance of our systems on the dataset. Overall, experimental results demonstrate the proposed duplex text normalization framework is highly effective and applicable to a range of domains and languages
Word embeddings are a powerful approach for analyzing language, and exponential family embeddings (EFE) extend them to other types of data. Here we develop structured exponential family embeddings (S-EFE), a method for discovering embeddings that var y across related groups of data. We study how the word usage of U.S. Congressional speeches varies across states and party affiliation, how words are used differently across sections of the ArXiv, and how the co-purchase patterns of groceries can vary across seasons. Key to the success of our method is that the groups share statistical information. We develop two sharing strategies: hierarchical modeling and amortization. We demonstrate the benefits of this approach in empirical studies of speeches, abstracts, and shopping baskets. We show how S-EFE enables group-specific interpretation of word usage, and outperforms EFE in predicting held-out data.
The recent advancement of pre-trained Transformer models has propelled the development of effective text mining models across various biomedical tasks. However, these models are primarily learned on the textual data and often lack the domain knowledg e of the entities to capture the context beyond the sentence. In this study, we introduced a novel framework that enables the model to learn multi-omnics biological information about entities (proteins) with the help of additional multi-modal cues like molecular structure. Towards this, rather developing modality-specific architectures, we devise a generalized and optimized graph based multi-modal learning mechanism that utilizes the GraphBERT model to encode the textual and molecular structure information and exploit the underlying features of various modalities to enable end-to-end learning. We evaluated our proposed method on ProteinProtein Interaction task from the biomedical corpus, where our proposed generalized approach is observed to be benefited by the additional domain-specific modality.
Traditional (unstructured) pruning methods for a Transformer model focus on regularizing the individual weights by penalizing them toward zero. In this work, we explore spectral-normalized identity priors (SNIP), a structured pruning approach that pe nalizes an entire residual module in a Transformer model toward an identity mapping. Our method identifies and discards unimportant non-linear mappings in the residual connections by applying a thresholding operator on the function norm. It is applicable to any structured module, including a single attention head, an entire attention block, or a feed-forward subnetwork. Furthermore, we introduce spectral normalization to stabilize the distribution of the post-activation values of the Transformer layers, further improving the pruning effectiveness of the proposed methodology. We conduct experiments with BERT on 5 GLUE benchmark tasks to demonstrate that SNIP achieves effective pruning results while maintaining comparable performance. Specifically, we improve the performance over the state-of-the-art by 0.5 to 1.0% on average at 50% compression ratio.

الأسئلة المقترحة

التعليقات
جاري جلب التعليقات جاري جلب التعليقات
سجل دخول لتتمكن من متابعة معايير البحث التي قمت باختيارها
mircosoft-partner

هل ترغب بارسال اشعارات عن اخر التحديثات في شمرا-اكاديميا