yuanxiaosc / Entity-Relation-Extraction

Entity and Relation Extraction Based on TensorFlow and BERT. 基于TensorFlow和BERT的管道式实体及关系抽取,2019语言与智能技术竞赛信息抽取任务解决方案。Schema based Knowledge Extraction, SKE 2019
https://yuanxiaosc.github.io/2019/05/17/多关系抽取研究/
1.22k stars 271 forks source link

您好,想知道run_sequnce_labeling代码中训练过程一开始的loss大概是多少? #40

Closed troublemaker-r closed 4 years ago

troublemaker-r commented 4 years ago

正在用pytorch复现,在这里遇了疑问,评价指标达不到您的效果,比较差。 loss = 0.5 * predicate_loss + token_label_loss 因为tensorflow看不到中间过程是具体输出,想请教您一下。

yuanxiaosc commented 4 years ago

可以用TensorFlow对应的tensorboard查看训练过程,目前TensorFlow 2 已经推出,建议你直接使用 https://github.com/huggingface/transformers 来构建