Tensorflow 2.0和较低版本的兼容问题
入坑TF较深,不同的算法需要版本不同,TF2.0向下不兼容,因此遇到很多问题
本人使用的是TF2.0
一、报错 AttributeError: module ‘tensorflow’ has no attribute ‘log’
解决: tf.math.log 代替tf.log
二、报错 AttributeError: module ‘tensorflow’ has no attribute ‘random_shuffle’
解决: tf.random.shuffle 代替tf.random_shuffle
三、报错 AttributeError:module ‘keras.engine.topology’ has no attribute 'load_weights_from_hdf5_group_by_name
解决: 参考 https://blog.csdn.net/c20081052/article/details/80745969
我使用第二种方法解决
四、报错:tf.reduce_mean got an unexpected keyword argument ‘keep_dims’
解决参考:https://blog.csdn.net/qq_36362060/article/details/80888949
五、报错::OperatorNotAllowedInGraphError: using a tf.Tensor as a Python bool is not allowed: AutoGraph is disabled in this function. Try decorating it directly with @tf.function.
————————————————
解决参考:开头添加 tf.compat.v1.enable_eager_execution()
六、报错:AttributeError: ‘Model’ object has no attribute ‘metrics_tensors’.
————————————————
解决参考:https://blog.csdn.net/u013066730/article/details/109071869
七、报错:AttributeError: ‘Model’ object has no attribute ‘_get_distribution_strategy’
————————————————
解决参考:https://blog.csdn.net/jiaxiaohui0409/article/details/109045857