任务简介--高性能小模型测评

这个任务的目标是训练一个和正常大小的语言模型效果相似的轻量级的语言模型。每个提交上来的模型都会在多个不同的下游NLP任务上评估性能。我们将会综合考虑模型参数数量，模型准确率以及模型推理时间，这些将一起作为模型的评估标准。为了满足很多参赛者对中文语料缺乏的情况，我们提供了目前为止最大中文语料库作为这个任务的补充资源。这些语料将会在之后公布给大家。

Contact: cluebenchmark@163.com
The goal of this task is to train a light language model which is still as powerful as the other normal models. Each model will be tested on many different downstream NLP tasks. We would take the number of parameters, accuracy and inference time as the metrics to measure the performance of a model. To meet the challenge of the lack of Chinese corpus, we will provide a big Chinese corpus for this task and will release them for all the researchers later.

github项目地址：https://github.com/CLUEbenchmark/LightLM
NLPCC2020的地址：https://mp.weixin.qq.com/s/FS2SFmP16IfUZsPp7F6LbA