Large Product Key Memory for Pretrained Language Models
arXiv:2010.03881
Abstract
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 We define a new memory usage metric, and careful observation using this metric reveals that most memory slots remain outdated during the training of PKM-augmented models. To train better PLMs by tackling this issue, we propose simple but effective solutions: (1) initialization from the model weights pretrained without memory and (2) augmenting PKM by addition rather than replacing a feed-forward network. We verify that both of them are crucial for the pretraining of PKM-augmented PLMs, enhancing memory utilization and downstream performance. Code and pretrained weights are available at https://github.com/clovaai/pkm-transformers.
arXiv commentAccepted to Findings of EMNLP 2020; 10 pages, 4 figures
References in corpus (17)
- Adam: A Method for Stochastic Optimization
- Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
- Distilling the Knowledge in a Neural Network
- Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
- Language Models are Few-Shot Learners
- Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism
- SQuAD: 100,000+ Questions for Machine Comprehension of Text
- REALM: Retrieval-Augmented Language Model Pre-Training
- NSML: Meet the MLaaS platform with a real-world case study
- NSML: A Machine Learning Platform That Enables You to Focus on Your Models
- Large Memory Layers with Product Keys
- Scaling Memory-Augmented Neural Networks with Sparse Reads and Writes
- Compressive Transformers for Long-Range Sequence Modelling
- Hierarchical Memory Networks
- Augmenting Self-attention with Persistent Memory
- Facts as Experts: Adaptable and Interpretable Neural Memory over Symbolic Knowledge
- Scheduled DropHead: A Regularization Method for Transformer Models