Shurui Gui

Hi @NarineK, I guess it's a good idea to wrap the model to make those inputs acceptable form to IG, because it only cares about inputs here. But for a...

> The version requirement for captum is **captum==0.2.0**. As version 0.5.0 is already published, is there any specific reason for the strong version requirement in DIG? In our code, some...

Thank you for your suggestion. My opinion is that it is the message passing nature to have directions, *e.g.*, the message passing from the base graph to motifs and the...

I also tried setting the type of dataset to a super class of Calendar: ```python def __init__(self, dataset: CalendarBase): ``` When I wrongly define the return_class type as: ```python class_from_function(Calendar.from_str,...

@mauvilsa Thank you for your workarounds! I would like to share one more user case: using [Equinox](https://github.com/patrick-kidger/equinox) in [JAX](https://github.com/google/jax), since every equinox Module is a dataclass.

This compatibility module is only available on dig-stable branch. For different dig, you may refer to the different tutorials. If you are unsure about the version you are using, I...

May you try to downgrade torch-geometric to 2.0.3? I remember this is an extra incompatibility problem from ~=2.0 to ~=2.1.

Hi @czstudio, Hope you are doing well. These figures are draw only in synthetic datasets with causal subgraph ground truths (nodes in your examples). The overall visualization strategy is using...

Sorry for the inconvenience. It sounds weird. Random seeds are reset in many places including the dataloader. Would you mind to 1) compare the model weights before training and 2)...

Hi Zhuangzhuang, The easiest implementation will be to modify the [`eval_score` function ](https://github.com/divelab/GOOD/blob/fff0e2dc311e1d0e0a23052f6c601db4f185e12e/GOOD/utils/evaluation.py#L60) so that it includes a new parameter for the selection of the score function like F1, Accuracy....