module ‘tokenization‘ has no attribute ‘FullTokenizer‘+‘Tensor‘ object has no attribute ‘numpy‘+tf.e

问题1：module ‘tokenization’ has no attribute ‘FullTokenizer’

导致这种问题有几种可能，关于tokennizetion

解决方法1：pip install bert-tensorflow==1.0.1
解决放法2：kaggle的话需要将tpkennizetion拷贝到working中
copyfile(src = “…/input/…/tokenization.py”, dst = “…/working/tokenization.py”)
import tokenization

问题二’Tensor’ object has no attribute ‘numpy’

根据你的tf版本选择不同的方式
查看tf版本

tf.__version__
tf.compat.v1.enable_eager_execution() # tf<1.4
tf.enable_eager_execution() # tensorflow >= 1.4 <2.0

问题三 tf.enable_eager_execution must be called at program startup.

重启kernal就可以