Amit Sharma

yeah, we do not support explaining regression models in pytorch whose output is a vector.

@mozolcer This might be because of the tf v1 compat code. Can you try the same code without the v1 compat lines? This error is often associated with lack of...

you may also try the latest version of DiCE, v0.9, that has updates to the tensorflow code and should resolve this bug.

thanks @elikling for adding this. Based on our discussion in the last meeting, do you think this information will be better as a docstring for the linear_dataset function? I'd encourage...

@daikikatsuragawa thanks for this contribution. Can you share a few lines on the motivation for this PR (e.g., why change `cfs` from [] to `pd.DataFrame()`?) and are there any bugs...

@AlxndrMlk this is intended behavior. To avoid the tricky task of auto-detecting the type of a variable, we are relying on the pandas dataframe to provide us the correct data...

We can consider adding a simple check, but the concern was computational efficiency. When I checked earlier, other libraries tended to loop through the entire data (O(N)) or check until...

@AlxndrMlk thanks for the thoughtful reply. Agree that we should be consistent with EconML as far as possible. > My proposal is that we add explicit checks to make sure...

yes, currently distance-matching and the three propensity based estimators check for binary treatment. Would make sense to modify the behavior for all such methods that do a binary treatment check.

This guide has some pointers to lint/code formatting checkers in PR checklist: https://github.com/py-why/dowhy/blob/main/docs/source/contributing/contributing-code.rst Other than that, I would suggest simply running all the tests. Since this change affects binary versus...