ConvBERT: Improving BERT with Span-based Dynamic Convolution

• URL: http://arxiv.org/abs/2008.02496v3
• Date: Tue, 2 Feb 2021 11:11:28 GMT
• Title: ConvBERT: Improving BERT with Span-based Dynamic Convolution
• Authors: Zihang Jiang, Weihao Yu, Daquan Zhou, Yunpeng Chen, Jiashi Feng, Shuicheng Yan
• Abstract summary: BERT heavily relies on the global self-attention block and thus suffers large memory footprint and computation cost. We propose a novel span-based dynamic convolution to replace these self-attention heads to directly model local dependencies. The novel convolution heads, together with the rest self-attention heads, form a new mixed attention block that is more efficient at both global and local context learning.
• Score: 144.25748617961082
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Pre-trained language models like BERT and its variants have recently achieved impressive performance in various natural language understanding tasks. However, BERT heavily relies on the global self-attention block and thus suffers large memory footprint and computation cost. Although all its attention heads query on the whole input sequence for generating the attention map from a global perspective, we observe some heads only need to learn local dependencies, which means the existence of computation redundancy. We therefore propose a novel span-based dynamic convolution to replace these self-attention heads to directly model local dependencies. The novel convolution heads, together with the rest self-attention heads, form a new mixed attention block that is more efficient at both global and local context learning. We equip BERT with this mixed attention design and build a ConvBERT model. Experiments have shown that ConvBERT significantly outperforms BERT and its variants in various downstream tasks, with lower training cost and fewer model parameters. Remarkably, ConvBERTbase model achieves 86.4 GLUE score, 0.7 higher than ELECTRAbase, while using less than 1/4 training cost. Code and pre-trained models will be released.

Related papers

• MosaicBERT: A Bidirectional Encoder Optimized for Fast Pretraining [10.421048804389343]
  We introduce MosaicBERT, a BERT-style encoder architecture and training recipe that is empirically optimized for fast pretraining. When pretrained from scratch on the C4 dataset, this base model achieves a downstream average GLUE (dev) score of 79.6 in 1.13 hours on 8 A100 80 GB GPUs at a cost of roughly $20. This empirical speed up in pretraining enables researchers and engineers to pretrain custom BERT-style models at low cost instead of finetune on existing generic models.
  arXiv  Detail & Related papers  (2023-12-29T06:05:19Z)
• MoEBERT: from BERT to Mixture-of-Experts via Importance-Guided Adaptation [68.30497162547768]
  We propose MoEBERT, which uses a Mixture-of-Experts structure to increase model capacity and inference speed. We validate the efficiency and effectiveness of MoEBERT on natural language understanding and question answering tasks.
  arXiv  Detail & Related papers  (2022-04-15T23:19:37Z)
• Deploying a BERT-based Query-Title Relevance Classifier in a Production System: a View from the Trenches [3.1219977244201056]
  Bidirectional Representations from Transformers (BERT) model has been radically improving the performance of many Natural Language Processing (NLP) tasks. It is challenging to scale BERT for low-latency and high- throughput industrial use cases due to its enormous size. We successfully optimize a Query-Title Relevance (QTR) classifier for deployment via a compact model, which we name BERT Bidirectional Long Short-Term Memory (BertBiLSTM) BertBiLSTM exceeds the off-the-shelf BERT model's performance in terms of accuracy and efficiency for the aforementioned real-world production task
  arXiv  Detail & Related papers  (2021-08-23T14:28:23Z)
• AutoBERT-Zero: Evolving BERT Backbone from Scratch [94.89102524181986]
  We propose an Operation-Priority Neural Architecture Search (OP-NAS) algorithm to automatically search for promising hybrid backbone architectures. We optimize both the search algorithm and evaluation of candidate models to boost the efficiency of our proposed OP-NAS. Experiments show that the searched architecture (named AutoBERT-Zero) significantly outperforms BERT and its variants of different model capacities in various downstream tasks.
  arXiv  Detail & Related papers  (2021-07-15T16:46:01Z)
• I-BERT: Integer-only BERT Quantization [78.43819756382103]
  We propose I-BERT, a novel quantization scheme for Transformer based models. I-BERT performs an end-to-end integer-only BERT inference without any floating point calculation. We show that for both cases, I-BERT achieves similar (and slightly higher) accuracy as compared to the full-precision baseline.
  arXiv  Detail & Related papers  (2021-01-05T02:42:58Z)
• DeBERTa: Decoding-enhanced BERT with Disentangled Attention [119.77305080520718]
  We propose a new model architecture DeBERTa that improves the BERT and RoBERTa models using two novel techniques. We show that these techniques significantly improve the efficiency of model pre-training and the performance of both natural language understanding (NLU) and natural langauge generation (NLG) downstream tasks.
  arXiv  Detail & Related papers  (2020-06-05T19:54:34Z)
• DynaBERT: Dynamic BERT with Adaptive Width and Depth [55.18269622415814]
  We propose a novel dynamic BERT model (abbreviated as DynaBERT) It can flexibly adjust the size and latency by selecting adaptive width and depth. It consistently outperforms existing BERT compression methods.
  arXiv  Detail & Related papers  (2020-04-08T15:06:28Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.