more information about build_input_head

[943fansi]

continuous_inputs, categorical_inputs = layers.build_input_head(window, continuous_columns, categorical_columns)

I want to use this function.build_input_head, but I don't know specific examples of continuous_columns and categorical_columns. I tried to use continuous_columns = ContinuousColumn(name='a', column_names=['a'], input_name='a') but got error AttributeError: 'str' object has no attribute 'input_dim'

[zhangxjohn]

Do you want to define a neural network model like the example https://github.com/DataCanvasIO/HyperTS/blob/main/examples/08_custom_search_space_02.ipynb?

[943fansi]

@zhangxjohn Thanks for your reply. Yes, I am currently studying the code in hyperts/framework/dl/models/deepar.py.

[zhangxjohn]

Ok, if you want to understand continuous_columns, then I suggest that you can debug the code (Details in function _preprocessor(self, X, y) of hyperts/framework/dl/_base.py). This is a more efficient way.Moreover, if you want to define a new neural network model, just replace backbone in example examples/08_custom_search_space_02.ipynb.

[943fansi]

Thank you for your suggestion, I am doing this. By the way, will the Transformer model or some updated models be added in the future? For example, from this repository Time-Series-Library.

[zhangxjohn]

Thank you for your suggestion, we are open to the idea. We also sincerely hope that you can contribute to the development of this project if you wish to do so.