Make models compatible to Captum #3990
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
rusty1s
reviewed
Feb 3, 2022
examples/captum_explainability.py
Outdated
| data = dataset[0] | ||
|
|
||
|
|
||
| class Net(torch.nn.Module): |
There was a problem hiding this comment.
Shall we use torch_geometric.nn.models.GCN here?
There was a problem hiding this comment.
I used the exact same model as in examples/gnn_explainer.py. Should I change it nevertheless?
There was a problem hiding this comment.
Yeah, I think it is better to directly use pre-defined models.
rusty1s
reviewed
Feb 4, 2022
| x = F.relu(self.conv1(x, edge_index)) | ||
| x = F.dropout(x, training=self.training) | ||
| x = self.conv2(x, edge_index) | ||
| return F.log_softmax(x, dim=1) |
There was a problem hiding this comment.
Ok, what does that mean in particular? Is the usage of log_softmax correct?
Co-authored-by: Matthias Fey <matthias.fey@tu-dortmund.de>
Co-authored-by: Matthias Fey <matthias.fey@tu-dortmund.de>
Co-authored-by: Matthias Fey <matthias.fey@tu-dortmund.de>
Co-authored-by: Matthias Fey <matthias.fey@tu-dortmund.de>
RBendias
changed the title
rusty1s
approved these changes
Feb 9, 2022
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
As discussed in issue #3858, we can add a
to_captummethod for simple use of Captum. The method adjusts the forward function of a model. I have added an example of edge explainability for node classification.I have tested several Captum methods (IntegratedGradients, Saliency, InputXGradient, GuidedBackprop, Deconvolution, ShapleyValueSampling, Kernelshap, FeatureAblation), all of which work appropriately with the Captum example given here.
@rusty1s I have implemented a
CaptumModelclass, theto_captummethod seems redundant. Alternatively, I could just overwrite the forward function instead of creating a new model. Do you have a preference and do you see an advantage in not creating a new model?