Language Models with Transformers

TL;DR

This paper introduces a new Transformer-based architecture with LSTM layers and a Coordinate Architecture Search method, significantly improving language modeling perplexity on standard datasets.

Contribution

It proposes a novel architecture combining Transformers and LSTMs, optimized via CAS, to enhance language modeling performance.

Findings

CAS achieves perplexities between 20.42 and 34.11

Improves perplexity by an average of 12.0 units over state-of-the-art LSTMs

Effective architecture for language modeling demonstrated on multiple datasets

Abstract

The Transformer architecture is superior to RNN-based models in computational efficiency. Recently, GPT and BERT demonstrate the efficacy of Transformer models on various NLP tasks using pre-trained language models on large-scale corpora. Surprisingly, these Transformer architectures are suboptimal for language model itself. Neither self-attention nor the positional encoding in the Transformer is able to efficiently incorporate the word-level sequential context crucial to language modeling. In this paper, we explore effective Transformer architectures for language model, including adding additional LSTM layers to better capture the sequential context while still keeping the computation efficient. We propose Coordinate Architecture Search (CAS) to find an effective architecture through iterative refinement of the model. Experimental results on the PTB, WikiText-2, and WikiText-103 show that CAS achieves perplexities between 20.42 and 34.11 on all problems, i.e. on average an improvement of 12.0 perplexity units compared to state-of-the-art LSTMs. The source code is publicly available.