I am trying to implement my own version of the lanczos algorithm. I would like to verify the output I am getting by checking against an easy to run and trustworthy implementation.

All I can find online returns the final eigendecomposition of a matrix rather than the tri-diagonal matrix. For example:

https://pypi.org/project/pylanczos/
https://github.com/mrcdr/lambda-lanczos

Neither of those seem to have a user facing documented cal to just give a matrix and obtain the tri-diagonalization.