ALBERT for Chinese Text Classification

albert-zh finetune for tensorflow 2.x

基于tf2及albert_zh的中文情感分析任务

requirements

tensorflow 2.3
python 3.8
albert-tiny-zh

##label

0: negative
1: positive

train

注意调整FLAGS配置

python run_classifier.py

infer

python predict_from_saved_model.py

example

"水果新鲜！发货快，服务好，京东物流顶呱呱，快递小哥服务好周到，送货上门，每次都热情满满，辛苦了，必须赞！",
"隔音差，加速有点不给力",
"最满意的是耗电量很低，百公里耗电才11.8度，这样算起来，出行成本比地铁的还低。"

output

cost time: 0.05300021171569824
['positive', 'negative', 'positive']

reference

https://github.com/brightmart/albert_zh

https://github.com/kpe/bert-for-tf2

https://github.com/kamalkraj/ALBERT-TF2.0