Tobias Pitters

Results 284 comments of Tobias Pitters

I looked into this and the problem here seems to be the order of sharding and batching/or how `drop_last_batch` is done (see the potential solutions below if unclear). Since we...

@FrenkT are you still working on this? I took a brief look and think we would need quite some changes here, e.g. your solution is a breaking change (you remove...

Shall we create tickets on other repositories or are there any other communication channels we have available with fellow maintainers?

Am I missing something here? On [prod](https://shap.readthedocs.io/en/latest/generated/shap.Explanation.html#shap.Explanation) I see: ![image](https://github.com/user-attachments/assets/16983078-eb17-44e8-a31c-9a5da42473bf) while on [this branch's docs](https://shap--3886.org.readthedocs.build/en/3886/generated/shap.Explanation.html#shap.Explanation) I just find: ![image](https://github.com/user-attachments/assets/3273ed8b-3df9-4684-8dbc-eb13655a57ca) I guess this branch is just a few commits behind. But...

> > Am I missing something here? > > I don't think so. The change in the rendering style is to be expected, as napolean and numpydoc render in slightly...

> We have a few options for type hints. At present we typically have the hints in two places, both in the function signature and also in the docstring: >...

Here are a couple things: - when multiple `Added/Changed in version x` appear, I think the old layout is nicer, since the new layout is IMO too dense, e.g. [here](https://shap--3886.org.readthedocs.build/en/3886/generated/shap.GPUTreeExplainer.html)...

> Thanks for the analysis! You spotted some interesting things that I'd missed. > > Yes, it's a shame the `versionadded` and `versionchanged` directives don't look as tidy. I'll see...

@venser12 Thanks for following this up. Could you please add a test for this? You could write the model to a temporary directory and load it back from there for...