| Autor: |
Agarwal, Naman, Chen, Xinyi, Dogariu, Evan, Feinberg, Vlad, Suo, Daniel, Bartlett, Peter, Hazan, Elad |
| Rok vydání: |
2024 |
| Předmět: |
|
| Druh dokumentu: |
Working Paper |
| Popis: |
We address the challenge of efficient auto-regressive generation in sequence prediction models by introducing FutureFill: a method for fast generation that applies to any sequence prediction algorithm based on convolutional operators. Our approach reduces the generation time requirement from linear to square root relative to the context length. Additionally, FutureFill requires a prefill cache sized only by the number of tokens generated, which is smaller than the cache requirements for standard convolutional and attention-based models. We validate our theoretical findings with experimental evidence demonstrating correctness and efficiency gains in a synthetic generation task. |
| Databáze: |
arXiv |
| Externí odkaz: |
|
