Character-Level Language Modeling with Deeper Self-Attention [article]

Rami Al-Rfou, Dokook Choe, Noah Constant, Mandy Guo, Llion Jones
<span title="2018-12-10">2018</span> <i > arXiv </i> &nbsp; <span class="release-stage" >pre-print</span>
LSTMs and other RNN variants have shown strong performance on character-level language modeling. These models are typically trained using truncated backpropagation through time, and it is common to assume that their success stems from their ability to remember long-term contexts. In this paper, we show that a deep (64-layer) transformer model with fixed context outperforms RNN variants by a large margin, achieving state of the art on two popular benchmarks: 1.13 bits per character on text8 and
more &raquo; ... .06 on enwik8. To get good results at this depth, we show that it is important to add auxiliary losses, both at intermediate network layers and intermediate sequence positions.
<span class="external-identifiers"> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/1808.04444v2">arXiv:1808.04444v2</a> <a target="_blank" rel="external noopener" href="https://fatcat.wiki/release/pzed56corracdi6vn3ltgz6vri">fatcat:pzed56corracdi6vn3ltgz6vri</a> </span>
<a target="_blank" rel="noopener" href="https://web.archive.org/web/20191027143031/https://arxiv.org/pdf/1808.04444v2.pdf" title="fulltext PDF download" data-goatcounter-click="serp-fulltext" data-goatcounter-title="serp-fulltext"> <button class="ui simple right pointing dropdown compact black labeled icon button serp-button"> <i class="icon ia-icon"></i> Web Archive [PDF] <div class="menu fulltext-thumbnail"> <img src="https://blobs.fatcat.wiki/thumbnail/pdf/95/a9/95a9ce8872576e0b7fe7ec02f178a26b5dbe6dcb.180px.jpg" alt="fulltext thumbnail" loading="lazy"> </div> </button> </a> <a target="_blank" rel="external noopener" href="https://arxiv.org/abs/1808.04444v2" title="arxiv.org access"> <button class="ui compact blue labeled icon button serp-button"> <i class="file alternate outline icon"></i> arxiv.org </button> </a>