DiffuSeq-v2: Bridging Discrete and Continuous Text Spaces for Accelerated Seq2Seq Diffusion Models

Shansan Gong, Mukai Li, Jiangtao Feng, Zhiyong Wu, Lingpeng Kong


Abstract
Diffusion models have gained prominence in generating high-quality sequences of text. Nevertheless, current approaches predominantly represent discrete text within a continuous diffusion space, which incurs substantial computational overhead during training and results in slower sampling speeds. In this paper, we introduce a soft absorbing state that facilitates the diffusion model in learning to reconstruct discrete mutations based on the underlying Gaussian space, thereby enhancing its capacity to recover conditional signals. During the sampling phase, we employ state-of-the-art ODE solvers within the continuous space to expedite the sampling process. Comprehensive experimental evaluations reveal that our proposed method effectively accelerates the training convergence by 4x and generates samples of similar quality 800x faster, rendering it significantly closer to practical application.
Anthology ID:
2023.findings-emnlp.660
Volume:
Findings of the Association for Computational Linguistics: EMNLP 2023
Month:
December
Year:
2023
Address:
Singapore
Editors:
Houda Bouamor, Juan Pino, Kalika Bali
Venue:
Findings
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
9868–9875
Language:
URL:
https://aclanthology.org/2023.findings-emnlp.660
DOI:
10.18653/v1/2023.findings-emnlp.660
Bibkey:
Cite (ACL):
Shansan Gong, Mukai Li, Jiangtao Feng, Zhiyong Wu, and Lingpeng Kong. 2023. DiffuSeq-v2: Bridging Discrete and Continuous Text Spaces for Accelerated Seq2Seq Diffusion Models. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 9868–9875, Singapore. Association for Computational Linguistics.
Cite (Informal):
DiffuSeq-v2: Bridging Discrete and Continuous Text Spaces for Accelerated Seq2Seq Diffusion Models (Gong et al., Findings 2023)
Copy Citation:
PDF:
https://preview.aclanthology.org/landing_page/2023.findings-emnlp.660.pdf