Solve valueerror: argument must be a deny tensor: 0 - got shape [198602], but wanted [198602, 16]

stay Tensorflow Call in tf.data.Dataset.from_generator(generator_function,args=(values1,values2))(generator_function Is a custom generator function ,args Is the input parameter of the generator ) The times made this mistake , All night long , It is actually an error reported in this position ：

stay tensor_util.py Inside make_tensor_proto() The following position of the function ：

Which is in the picture nparray = np.array(values, dtype=np_dt), This values Is the input parameter of the generator values1,values2（ How many parameters will judge how many times ,values=values1,values=values2）, Above picture _GetDenseDimensions(values) The function is defined as follows ：

It can be found through pre and post analysis , This args The parameters inside cannot be the list of the list , Can be a list , It can also be strings and so on , However, once the list contains the parameters of the list, an error will be reported .