Salvato in:
| Autori principali: | , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2022
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2204.05424 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866913245560832000 |
|---|---|
| author | Kasai, Jungo Sakaguchi, Keisuke Bras, Ronan Le Radev, Dragomir Choi, Yejin Smith, Noah A. |
| author_facet | Kasai, Jungo Sakaguchi, Keisuke Bras, Ronan Le Radev, Dragomir Choi, Yejin Smith, Noah A. |
| contents | Text generation with beam search has proven successful in a wide range of applications. We point out that, though largely overlooked in the literature, the commonly-used implementation of beam decoding (e.g., Hugging Face Transformers and fairseq) uses a first come, first served heuristic: it keeps a set of already completed sequences over time steps and stops when the size of this set reaches the beam size. Based on this finding, we introduce a patience factor, a simple modification to this beam decoding implementation, that generalizes the stopping criterion and provides flexibility to the depth of search. Empirical results demonstrate that adjusting this patience factor improves decoding performance of strong pretrained models on news text summarization and machine translation over diverse language pairs, with a negligible inference slowdown. Our approach only modifies one line of code and can be thus readily incorporated in any implementation. Further, we find that different versions of beam decoding result in large performance differences in summarization, demonstrating the need for clarity in specifying the beam search implementation in research work. Our code will be available upon publication. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2204_05424 |
| institution | arXiv |
| publishDate | 2022 |
| record_format | arxiv |
| spellingShingle | A Call for Clarity in Beam Search: How It Works and When It Stops Kasai, Jungo Sakaguchi, Keisuke Bras, Ronan Le Radev, Dragomir Choi, Yejin Smith, Noah A. Computation and Language Text generation with beam search has proven successful in a wide range of applications. We point out that, though largely overlooked in the literature, the commonly-used implementation of beam decoding (e.g., Hugging Face Transformers and fairseq) uses a first come, first served heuristic: it keeps a set of already completed sequences over time steps and stops when the size of this set reaches the beam size. Based on this finding, we introduce a patience factor, a simple modification to this beam decoding implementation, that generalizes the stopping criterion and provides flexibility to the depth of search. Empirical results demonstrate that adjusting this patience factor improves decoding performance of strong pretrained models on news text summarization and machine translation over diverse language pairs, with a negligible inference slowdown. Our approach only modifies one line of code and can be thus readily incorporated in any implementation. Further, we find that different versions of beam decoding result in large performance differences in summarization, demonstrating the need for clarity in specifying the beam search implementation in research work. Our code will be available upon publication. |
| title | A Call for Clarity in Beam Search: How It Works and When It Stops |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2204.05424 |