arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2106.04563cs.CLcs.AIcs.LG
收藏

XtremeDistilTransformers:任务无关蒸馏的任务转移

XtremeDistilTransformers: Task Transfer for Task-agnostic Distillation

Microsoft Research(微软研究院)

Subhabrata Mukherjee, Ahmed Hassan Awadallah, Jianfeng Gao

摘要:

虽然深度和大型的预训练模型是各种自然语言处理任务的最新技术,但它们的巨大规模对资源受限环境中的实际使用提出了重大挑战。最近的工作在知识蒸馏提出任务不可知的以及特定于任务的方法来压缩这些模型,与特定于任务的往往会产生更高的压缩率。在这项工作中,我们开发了一个新的任务不可知的蒸馏框架XtremeDistilTransformers,利用特定于任务的方法来学习一个小型的通用模型,可以应用于任意任务和语言的优势。为此,我们研究了几个源任务的可移植性,增强资源和模型架构的蒸馏。我们评估了我们的模型在多个任务上的性能,包括通用语言理解评估(GLUE)基准,SQuAD问答数据集和包含41种语言的大规模多语言NER数据集。我们发布了三个与任务无关的检查点,分别为13MM,22MM和33MM参数,在多个任务中获得SOTA性能。

英文摘要:

While deep and large pre-trained models are the state-of-the-art for various natural language processing tasks, their huge size poses significant challenges for practical uses in resource constrained settings. Recent works in knowledge distillation propose task-agnostic as well as task-specific methods to compress these models, with task-specific ones often yielding higher compression rate. In this work, we develop a new task-agnostic distillation framework XtremeDistilTransformers that leverages the advantage of task-specific methods for learning a small universal model that can be applied to arbitrary tasks and languages. To this end, we study the transferability of several source tasks, augmentation resources and model architecture for distillation. We evaluate our model performance on multiple tasks, including the General Language Understanding Evaluation (GLUE) benchmark, SQuAD question answering dataset and a massive multi-lingual NER dataset with 41 languages. We release three distilled task-agnostic checkpoints with 13MM, 22MM and 33MM parameters obtaining SOTA performance in several tasks.

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。

↑