arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2403.02249cs.CVcs.AI
收藏

非自回归序列到序列视觉-语言模型

Non-autoregressive Sequence-to-Sequence Vision-Language Models

AWS AI Labs(AWS AI实验室)

Kunyu Shi, Qi Dong, Luis Goncalves, Zhuowen Tu, Stefano Soatto

AI总结:

针对自回归视觉-语言模型推理延迟高的问题,提出基于 Query-CTC 损失和并行解码的 NARVL,以联合 token 分布建模实现恒定时间推理,并保持与最先进自回归模型相当的性能。

摘要:

序列到序列视觉-语言模型正展现出良好前景,但由于其以自回归方式生成预测,推理延迟限制了适用性。我们提出一种并行解码的序列到序列视觉-语言模型,使用 Query-CTC 损失进行训练,该损失在解码器中对多条推理路径进行边缘化。这使我们能够建模 token 的联合分布,而不是像自回归模型那样局限于条件分布。所得模型 NARVL 达到了与其最先进自回归对应模型相当的性能,同时在推理时更快,将顺序生成 token 所带来的线性复杂度降低为恒定时间联合推理范式。

英文摘要:

Sequence-to-sequence vision-language models are showing promise, but their applicability is limited by their inference latency due to their autoregressive way of generating predictions. We propose a parallel decoding sequence-to-sequence vision-language model, trained with a Query-CTC loss, that marginalizes over multiple inference paths in the decoder. This allows us to model the joint distribution of tokens, rather than restricting to conditional distribution as in an autoregressive model. The resulting model, NARVL, achieves performance on-par with its state-of-the-art autoregressive counterpart, but is faster at inference time, reducing from the linear complexity associated with the sequential generation of tokens to a paradigm of constant time joint inference.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑