Overcoming Library Dependency Challenges When Installing CausalML

[shakednave-vi]

I need to install causalml, but it requires me to use sklearn==1.2.2 and pandas==1.3.5 in my pip virtual environment to avoid conflicts. Is there a solution to this issue caused by causalml’s outdated library dependencies?

[jeongyoonlee]

Hi @shakednave-vi.

We relaxed the pandas (and numpy) version requirement in #681, but it's not trivial for scikit-learn as our causal tree implementation depends on a specific version of scikit-learn.

cc: @alexander-pv

[shakednave-vi]

Hi @jeongyoonlee,

Can you please refer to the non-trivial parts in the package and I'll see if I can contribute?

[jeongyoonlee]

It involves removing dependencies on the Scikit-learn's tree module within causalml.inference.tree.causal.*. Check out #686 for more details and one proposal by @alexander-pv.

[jeongyoonlee]

I'm closing this issue as the pandas version requirement has been relaxed in #681, and we have #682, which is more specific for the scikit-learn dependency.