Early stopping conditioned on metric `val_loss` which is not available. Available metrics are: loss,

问题：

tensorflow:Early stopping conditioned on metric val_loss which is not available. Available metrics are: loss,accuracy

解决办法

：
增加验证集大小，即可看见训练过程中loss，accuracy，val_loss和val_accuracy。

提醒：

笔者train_generator、test_generator、val_generator都是通过TFRecord把文件夹中分配好的数据集读取并按文件夹分类的categorical labels；

model.compile(loss='categorical_crossentropy',optimizer='adam', metrics=[