Aobo Yang

Unfortunately, as you may have noticed, PyTorch ecosystem, or even the core torch lib itself, does not have a standardized way of writing docstring. For example, [the docs of cnn](https://pytorch.org/docs/stable/generated/torch.nn.Conv3d.html#torch.nn.Conv3d)...

I have built the website and test this change. Unfortunately, `autodoc_preserve_defaults` breaks our style. The style issue may be because :obj is for code piece so it inserts an extra...

@ProGamerGov This PR has contained too many contents for me to directly review. Could you help me break down things a little bit. Based on my understanding, the changes can...

Great! If the `tensor` type thing is not too complicated, let's include it in this pr.

@CleonWong @caesar-one we have prepared a tutorial. You can find it at https://captum.ai/tutorials/Image_and_Text_Classification_LIME

hi @jlubao , as @felixmeyjr said, `captum.influence` is a new function introduced in the latest version 0.5.0. You may refer to our change logs for details https://github.com/pytorch/captum/releases

hi @ik362 , sorry for my late reply. I believe `pytorch-lightning` has nothing to do here. It will work as long as your model have a `forward`-like interface to pass...

Hi, thank you for making this, but I may miss some context/history here. Why do we need the "composable loss" in Captum? Pytorch has already provided a convention for loss:...

Could you elaborate more on this > I get attribution scores that are almost always the last few tokens of each sequence Do you mean you can get the attribution...

@vignesh28121997 do you mind elaborating a little bit more about your use case? what `capture` do you need to use in code?