我在训练集中加入了两种大概500个左右的新实体类别，在训练完毕后输出的新类别识别率均为0，查看预测结果同样对于新实体类别没有识别出来，我在label_list中添加了新类别，同时修改了训练文件，能否咨询一下该模型是否支持加入新的实体类别

您好！我在进行模型训练时遇到了这个问题：
tensorflow.python.framework.errors_impl.DataLossError: Unable to open table file C:\Users\Desktop\ner_zh\bert_ner\bert_ner\chinese_L-12_H-768_A-12\bert_model.ckpt: Data loss: not an sstable (bad magic number): perhaps your file is in a different file format and you need to use a different restore operator?
请问该如何正确加载模型呢？试过将"bert_model.ckpt.data-00000-of-00001"的后缀删除但并不好用。期待您的回复！

我运行出现错误提示如下（使用的是tensorflow1.13.1）：
No such file or directory: 'D:\Workproject\bert_ner\bert_ner\output\label2id.pkl'

请问怎么解决？

完整的输出信息如下：

Namespace(batch_size=32, bert_config_file='D:\Workproject\bert_ner\bert_ner\chinese_L-12_H-768_A-12\bert_config.json', bert_path='D:\Workproject\bert_ner\bert_ner\chinese_L-12_H-768_A-12', cell='lstm', clean=True, clip=0.5, data_dir='D:\Workproject\bert_ner\bert_ner\NERdata', device_map='0', do_eval=True, do_lower_case=True, do_predict=True, do_train=True, dropout_rate=0.5, filter_adam_var=False, init_checkpoint='D:\Workproject\bert_ner\bert_ner\chinese_L-12_H-768_A-12\bert_model.ckpt', label_list=None, learning_rate=1e-05, lstm_size=128, max_seq_length=512, model_path='D:\Workproject\bert_ner\bert_ner\output', ner='ner', num_layers=1, num_train_epochs=30, output_dir='D:\Workproject\bert_ner\bert_ner\output2', save_checkpoints_steps=200, save_summary_steps=200, verbose=False, vocab_file='D:\Workproject\bert_ner\bert_ner\chinese_L-12_H-768_A-12\vocab.txt', warmup_proportion=0.1)
2020-12-05 15:45:05.235899: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2
checkpoint path:D:\Workproject\bert_ner\bert_ner\output\checkpoint
Traceback (most recent call last):
File "d:/Workproject/bert_ner/bert_ner_api.py", line 50, in
predict_cmd()
File "d:/Workproject/bert_ner/bert_ner_api.py", line 28, in predict_cmd
from src.bert_ner_predict import ner_predict
File "d:\Workproject\bert_ner\src\bert_ner_predict.py", line 49, in
with codecs.open(os.path.join(model_dir, 'label2id.pkl'), 'rb') as rf:
File "C:\Users\hxuwjj\AppData\Local\Programs\Python\Python36\lib\codecs.py", line 897, in open
file = builtins.open(filename, mode, buffering)
FileNotFoundError: [Errno 2] No such file or directory: 'D:\Workproject\bert_ner\bert_ner\output\label2id.pkl'

你好，请问如何使用多GPU训练，感谢！

我在“AI前线“的公众号上看到 BertNER 的预测速度可以达到 80ms，想问一下您是如何计算的，有用到一些模型的优化技术吗？我使用 Tensorflow serving 在 P100 下的预测速度都只能达到 400ms 左右。希望能得到您的指导~

我是在网上看到《基于Bert-NER构建特定领域的中文信息抽取框架（上）》，感觉你们做的东西很好，前来学习一下。不知关系抽取的代码有公布出来了吗？
同时，发现文章中提到关系抽取是在《基于Bert-NER构建特定领域的中文信息抽取框架（下）》中详细介绍，但并没找到这个“下篇”。能否提供一下？
谢谢！

your code are very concise and clear to read. thanks a lot!!

could you tell me what's tensorflow version your code support?
I hope to reproduce your code. thanks again

程序运行到这里的时候，就会被Killed掉
tf.estimator.train_and_evaluate(estimator, train_spec, eval_spec)
想知道是为什么吗？

你好，我用tensorflow版本为1.14.0，运行你的代码出现
Traceback (most recent call last):
File "bert_ner_api.py", line 47, in
train_api()
File "bert_ner_api.py", line 21, in train_api
train(args)
File "/home/desktop/KG/NER-Chinese/bert_ner/bert_ner/bert_base/train/bert_lstm_ner.py", line 677, in train
early_stopping_hook = tf.contrib.estimator.stop_if_no_decrease_hook(
AttributeError: module 'tensorflow.contrib.estimator' has no attribute 'stop_if_no_decrease_hook'
麻烦帮解决一下

To use the CPU if you have installed tensorflow, or use the GPU if you have installed tensorflow-gpu:

python3.5 main.py

请问这个main.py文件在Neuro-master的哪个文件夹路径下呀？
(NeroNER) zhao@cumt:~/anaconda3/envs/NeroNER/NeuroNER-master$ ls
data LICENSE MANIFEST.in neuroner parameters.ini README.md requirements.txt setup.py test