2003
Search results
A hybrid approach to retrieval augmentation makes the most of your compute. M De Jong, Y Zemlyanskiy, N FitzGerald, J Ainslie, S Sanghai, F Sha, ... International Conference on Machine...
Jan 25, 2023 · View a PDF of the paper titled Pre-computed memory or on-the-fly encoding? A hybrid approach to retrieval augmentation makes the most of your compute, by Michiel de Jong and 6 other authors
- arXiv:2301.10448 [cs.CL]
- ICML 2023
Dec 23, 2023 · Autoregressive decoder inference is a severe bottleneck for Transformer models due to the memory bandwidth overhead from loading decoder weights and all attention keys and values at every decoding step (Shazeer, 2019; Pope et al., 2022; de Jong et al., 2022).
Oct 12, 2021 · Michiel de Jong, Yury Zemlyanskiy, Nicholas FitzGerald, Fei Sha, William Cohen. View a PDF of the paper titled Mention Memory: incorporating textual knowledge into Transformers through entity mention attention, by Michiel de Jong and 4 other authors.
- Michiel de Jong, Yury Zemlyanskiy, Nicholas FitzGerald, Fei Sha, William Cohen
- arXiv:2110.06176 [cs.CL]
- 2021
Michiel de Jong was born on 17 May 1973 in Oranjestad, Aruba. He is an actor and director, known for Black Book (2006), Julia's Tango (2007) and Het huis Anubis (2006).
- January 1, 1
- 3 min
- Oranjestad, Aruba
no code implementations • 17 Jun 2023 • Michiel de Jong, Yury Zemlyanskiy, Nicholas FitzGerald, Sumit Sanghai, William W. Cohen, Joshua Ainslie Memory-augmentation is a powerful approach for efficiently incorporating external information into language models, but leads to reduced performance relative to retrieving text.
Michiel de Jong, Vera Stara, Viviane von Döllen, Daniel Bolliger, Marcel Heerink, Vanessa Evers: Users requirements in the design of a virtual agent for patients with dementia and their caregivers. GOODTECHS 2018: 136-141.
