Efficient Transformers with Dynamic Token Pooling

Add to your list(s) Download to your calendar using vCal

• Piotr Nawrot, University of Edinburgh
• Thursday 02 March 2023, 11:00-12:00
• GR04, English Faculty Building, 9 West Road, Sidgwick Site.

If you have a question about this talk, please contact Panagiotis Fytas.

Transformers achieve unrivalled performance in modelling language, but remain inefficient in terms of memory and time complexity. A possible remedy is to reduce the sequence length in the intermediate layers by pooling fixed-length segments of tokens. Nevertheless, natural units of meaning, such as words or phrases, display varying sizes. To address this mismatch, we equip language models with a dynamic-pooling mechanism, which predicts segment boundaries in an autoregressive fashion. We compare several methods to infer boundaries, including end-to-end learning through stochastic re-parameterisation, supervised learning (based on segmentations from subword tokenizers or spikes in conditional entropy), as well as linguistically motivated boundaries. We perform character-level evaluation on texts from multiple datasets and morphologically diverse languages. The results demonstrate that dynamic pooling, which jointly segments and models language, is often both faster and more accurate than vanilla Transformers and fixed-length pooling within the same computational budget.

This talk is part of the Language Technology Lab Seminars series.

This talk is included in these lists:

• Cambridge Centre for Data-Driven Discovery (C2D3)
• Cambridge Forum of Science and Humanities
• Cambridge Language Sciences
• Cambridge talks
• Chris Davis' list
• GR04, English Faculty Building, 9 West Road, Sidgwick Site
• Guy Emerson's list
• Interested Talks
• Language Sciences for Graduate Students
• Language Technology Lab Seminars
• Simon Baker's List
• Trust & Technology Initiative - interesting events
• bld31
• ndk22's list
• ob366-ai4er
• rp587

Note that ex-directory lists are not shown.