Issues
- 0
Potential issue with learning BatchNorm parameters
#204 opened by MikiFER - 3
ResNet: Unstable Attributions
#194 opened by chr5tphr - 5
Model inference outside of composite context
#200 opened by MikiFER - 2
Changing The Value of epsilon in the composite Epsilonplusflat creates a bug
#203 opened by StudentAlessandro - 2
- 0
- 2
- 0
Core/Composite: Simplify Composite.context
#156 opened by chr5tphr - 2
Module with Multiple Inputs
#176 opened by rachtibat - 0
Core: GradOutHooks
#141 opened by chr5tphr - 0
Core: Warn the user when falling back to gradient
#147 opened by chr5tphr - 12
Numerical instability in ResNet50 heatmaps
#148 opened by rodrigobdz - 1
Custom rules for vision transformer
#184 opened by ascdqz - 1
LinearAttention Module
#169 opened by rachtibat - 2
How to add support to DenseNet from torchvision
#165 opened by mstaczek - 6
Layer-wise LRP score
#167 opened by shivmgg - 0
Canonizers: BatchNorm loses bias relevance
#193 opened by chr5tphr - 2
Smooth MaxPool2D rule
#181 opened by rachtibat - 0
- 29
- 12
- 2
Feature Request: Debugger
#187 opened by rachtibat - 4
Bug: MultiheadAttention in types
#188 opened by rachtibat - 1
- 3
Composite: bug in BetaSmooth constructor?
#172 opened by annahdo - 7
error occurred when run tutorial on custom model
#170 opened by ascdqz - 4
Core: Second Order Gradients
#142 opened by chr5tphr - 1
- 2
Support for padding layers
#154 opened by Samsdas - 8
Notebook for feed_forward.py
#24 opened by p16i - 2
Bug in AlphaBeta rule?
#107 opened by nkoenen - 2
Feature request: Assign rule by layer index
#76 opened by rodrigobdz - 2
- 2
Canonizers
#63 opened by rachtibat - 2
torch 1.10 version not working
#60 opened by rachtibat - 0
RuntimeError: Mismatch in shape: grad_output[0] has a shape of torch.Size([1, 1, 80, 80]) and output[0] has a shape of torch.Size([1, 1, 96, 96]).
#59 opened by bugsuse - 0
RuntimeError: Mismatch in shape: grad_output[0] has a shape of torch.Size([1, 1, 80, 80]) and output[0] has a shape of torch.Size([1, 1, 96, 96]).
#58 opened by bugsuse - 1
Alpha Beta still includes Bias two times
#53 opened by rachtibat - 2
Default Rule
#51 opened by rachtibat - 0
- 4
- 2
[How to use] heatmap variants [...] such as the guided_backprop or smoothgrad [...] ?
#40 opened by sebastian-lapuschkin - 0
feed_forward is broken for torchvision 0.10.0
#37 opened by chr5tphr - 4
- 0
Image fetching in example is broken
#26 opened by chr5tphr - 4
Flat Rule for Pooling Layers
#9 opened by rachtibat - 2
Memory Leak
#11 opened by rachtibat