What does the VarLengthMultiheadSA do to SparseTrTensor input

[dattnguyenx]

Hi, the project is very helpful. thank you. I'd like to know what the operations are in the VarLengthMultiheadSA and how to utilize this module to build a network. Could you provide an example? the two projects you mentioned do not use this function to create a transformer-based network. Thank you again.