| Challenge: | Large Transformer-based models are reduced to a smaller number of self-attention heads and layers. |
| Approach: | They propose to prune BERT self-attention heads and layers to find subnetworks with comparable performance . they also extend this technique to multi-layer perceptrons to find out if they are unstable . |
| Outcome: | The proposed models are able to achieve 90% of full model performance with structured pruning and similar-sized subnetworks sampled from the rest of the model perform worse. |
Similar Papers
Learning to Win Lottery Tickets in BERT Transfer via Task-agnostic Mask Training (2022.naacl-main)
Copied to clipboard
| Challenge: | Recent studies show pre-trained language models contain matching subnetworks that have similar transfer learning performance as the original PLM. |
| Approach: | They propose to prune matching subnetworks using magnitude-based pruning . they propose to optimize the subnetwork structure towards the pre-training objectives . |
| Outcome: | The proposed method is more efficient in searching subnetworks and advantageous when fine-tuning within a range of data scarcity. |
Less Is More: Domain Adaptation with Lottery Ticket for Reading Comprehension (2021.findings-emnlp)
Copied to clipboard
| Challenge: | Existing domain adaptation paradigms for reading comprehension require large amounts of annotation data to achieve the desired task performance. |
| Approach: | They propose a few-shot domain adaptation paradigm for reading comprehension . they introduce self-attention attribution to weigh parameters and refine the lottery subnetwork . |
| Outcome: | The proposed model outperforms the full model fine-tuning adaptation on four out of five domains with a small amount of data available for adaptation. |
Finding the Dominant Winning Ticket in Pre-Trained Language Models (2022.findings-acl)
Copied to clipboard
| Challenge: | Existing studies on pre-trained language models show that they can fine-tune parameters but achieve good downstream performance. |
| Approach: | They find that a dominant winning ticket takes up 0.05% of the parameters and is transferable across different tasks. |
| Outcome: | The proposed model can achieve comparable performance with the full-parameter model, the authors show . the dominant winning ticket takes up 0.05% of the parameters, and the model is transferable across tasks, they show - the authors conclude . |
Revealing the Dark Secrets of BERT (D19-1)
Copied to clipboard
| Challenge: | Existing models of BERT-based learning systems are lacking specific mechanisms that contribute to its success. |
| Approach: | They propose to use GLUE tasks to analyze the interpretation of self-attention, which is one of the underlying components of BERT. |
| Outcome: | The proposed model outperforms the regular model on GLUE tasks by disabling attention in certain heads. |
Losing Heads in the Lottery: Pruning Transformer Attention in Neural Machine Translation (2020.emnlp-main)
Copied to clipboard
| Challenge: | Recent research shows that attention heads are not confident in their decisions and can be pruned. |
| Approach: | They apply the lottery ticket hypothesis to prune heads in early training . they find that the pruned model is 1.5 times faster at inference . |
| Outcome: | The proposed method is 1.5 times faster at inference, but at the cost of longer training. |
Successfully Applying the Stabilized Lottery Ticket Hypothesis to the Transformer Architecture (2020.acl-main)
Copied to clipboard
| Challenge: | Current neural networks are heavily growing in depth, with many fully connected layers. |
| Approach: | They propose to combine stabilized lottery ticket pruning with unstructured pruning to improve model performance. |
| Outcome: | The proposed pruning techniques outperform all other techniques for even higher sparsity levels. |
EarlyBERT: Efficient BERT Training via Early-bird Lottery Tickets (2021.acl-long)
Copied to clipboard
| Challenge: | Large-scale pre-trained language models require enormous computational resources and long training time. |
| Approach: | They propose an algorithm to reduce inference time and train large NLP models by slimming the self-attention and fully-connected sub-layers inside a transformer. |
| Outcome: | The proposed algorithm achieves comparable performance to standard BERT with 35 45% less training time. |
Super Tickets in Pre-Trained Language Models: From Model Compression to Improving Generalization (2021.acl-long)
Copied to clipboard
Chen Liang, Simiao Zuo, Minshuo Chen, Haoming Jiang, Xiaodong Liu, Pengcheng He, Tuo Zhao, Weizhu Chen
| Challenge: | 'lottery tickets' can be trained to match the performance of a full model . subnetwork training can also outperform random sampled subnetworks of the same size . |
| Approach: | They propose to train a subnetwork of 'lottery tickets' to match the full model's performance. |
| Outcome: | The proposed model outperforms subnetworks of the same size in a phase transition phenomenon . the proposed model improves single task fine-tuning by 0.9 points on BERT-base and 1.0 points on GLUE large . |
Telling BERT’s Full Story: from Local Attention to Global Aggregation (2021.eacl-main)
Copied to clipboard
| Challenge: | Recent work discouraging the use of attention distributions for explaining a model’s behaviour suggests that attention distribution can provide insights into local behaviour of attention heads. |
| Approach: | They propose a distinction between local patterns revealed by attention and global patterns that refer back to the input and analyze BERT from both angles. |
| Outcome: | The proposed model can explain local behaviour of attention heads by comparing local and global patterns from both angles. |
Rethinking Network Pruning – under the Pre-train and Fine-tune Paradigm (2021.naacl-main)
Copied to clipboard
| Challenge: | Existing pruning results on benchmark transformers, such as BERT, are not as remarkable as those of convolutional neural networks. |
| Approach: | They propose to apply a knowledge-aware pruning process to transformer-based pre-trained language models to reduce model size and model weight. |
| Outcome: | The proposed pruning method outperforms the leading competitors with a 20-times weight/FLOPs compression and neglectable loss in prediction accuracy. |