arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2108.03340cs.CLcs.SDeess.AS
收藏

Seq2Seq的微型神经模型

Tiny Neural Models for Seq2Seq

Google Research(谷歌研究院)

Arun Kandoor

摘要:

具有面向任务的对话系统中的应用的语义解析模型需要在设备上运行高效的序列到序列(seq 2seq)架构。为此,我们提出了一种基于投影的编码器-解码器模型,称为pQRNN-MAtt。基于投影方法的研究仅限于编码器模型,我们相信这是第一个将其扩展到seq 2seq架构的研究。由此产生的量化模型大小小于3.5MB,非常适合设备上延迟关键应用。我们表明,在MTOP(一个具有挑战性的多语言语义解析数据集)上,模型的平均性能超过了基于LSTM的seq 2 seq模型,该模型使用预训练的嵌入,尽管模型小了85倍。此外,该模型可以有效地提取大型预训练模型,如T5/BERT。

英文摘要:

Semantic parsing models with applications in task oriented dialog systems require efficient sequence to sequence (seq2seq) architectures to be run on-device. To this end, we propose a projection based encoder-decoder model referred to as pQRNN-MAtt. Studies based on projection methods were restricted to encoder-only models, and we believe this is the first study extending it to seq2seq architectures. The resulting quantized models are less than 3.5MB in size and are well suited for on-device latency critical applications. We show that on MTOP, a challenging multilingual semantic parsing dataset, the average model performance surpasses LSTM based seq2seq model that uses pre-trained embeddings despite being 85x smaller. Furthermore, the model can be an effective student for distilling large pre-trained models such as T5/BERT.

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。

↑