How to use tabnet model to train models with both categorical features and continuous features?

[MrPolong]

I want to use my data to train tabnetclassfier models. The data is a hybrid data, including categorical features and continuous features. I find 'train_embedding.py' and 'train_iris.py' in /examples, which are use their own data types. I handle continuous data use tf.feature_column.numeric_column(col_name), and handle categorical data use tf.feature_column.indicator_column(tf.feature_column.categorical_column_with_vocabulary_list(col_name,[1,2,10,20,30])), Then I put them in feature_columns as a parameter of TabNetClassifier(). The console will raise the error of unequal size, such as

ValueError: Dimensions must be equal, but are 89 and 458 for '{{node tab_net_classifier_2/tab_net_2/Mul_10}} = Mul[T=DT_FLOAT](tab_net_classifier_2/tab_net_2/PartitionedCall, tab_net_classifier_2/tab_net_2/input_gn/Reshape_3)' with input shapes: [?,89], [?,458]

So, how do I train the model with two kinds of data as input.

[tchangaghis]

hello @MrPolong did you find a solution? i am looking also for something like that. thanks