How to handle variable length in the combination of Deep learning methods architecture

I am working on a variable length classification problem. I want to utilize multiple Deep learning methods in combination, like CNN, LSTM, attention, etc. Now I am pretty confused and struggling in feeding the data into the model and getting an output. I was getting multiple errors but could not get the complete idea of why they were coming.
Therefore, I am eagerly looking for help from where I can get implementation help. Any resources, or any guidance? It would be highly appreciated.

maybe this: düzensiz tensörler  |  TensorFlow Core
can help you get started

@Marcus : I don’t know why the chinese characters in the link above but the link is a regular one to tensorflow.org