Daxiong

icon indicating an email [email protected]

icon company ccnu icon location china icon location I am interested in NLP and Education Data Mining

Results 5 comments of Daxiong

请教复现问题

> 没看出来什么问题, 是否没有贴全 训练得acc 一直很低

请问，lac Andoroid的2m的小模型是怎么得到的？有没有转换成nb格式之前的模型吗

> > _No description provided._ > > Android版本的小模型经过PaddleLite优化得到，具体可参考： > https://paddle-lite.readthedocs.io/zh/latest/user_guides/model_optimize_tool.html > > 转换之前的模型，请参考release页面的相关下载： > https://github.com/baidu/lac/releases/ 转换之前的6m的模型我利用paddlelite的python接口转化成nb文件，模型的大小没有变化，所以我能理解为：通过您提供的第一个链接中的接口可以完成有效的压缩/

[Android] Crashing when sending a message

> Same here (model is downloaded in local successfully, but crash when sending the first message), my device is Huawei P30 Pro with Kirin 980 chipset, 8G RAM. > >...

PPO采用zero 3 stage后产生time out error

> generate_max_len 遇到一样的问题，改成2之后正常

this method can be used to bloom model?

> It is general for Transformer based large language models. We evaluate on LLaMA mostly in our paper because of its superior performance. We have additional results on Pythia and...