captum
captum copied to clipboard
pytorch
Model interpretability and understanding for PyTorch
## ❓ Questions and Help Hi Authors, When I apply LayerIntegratedGradients on Llama for long context QA, the feature attribution seems not meaningful. For example, when I copy a news...
(making a new PR since I missed creating the design doc earlier) RE: #1246 # IDGI API Design ## Background IDGI (Important Direction Integrated Gradients) is a generalized framework that...
## ❓ Questions and Help Hi! I'm facing a problem: We want to use Integrated gradient(IG) to explain a GCN model. We want to generate edge_mask for each edge in...
## 🐛 Bug Hi I am using captum in a project of mine as the main implantation source of the attribution methods. I am using the ResNet50 model that was...
## 🐛 Bug `AssertionError: Tensor target dimension torch.Size([1, 1, 224, 224]) is not valid. torch.Size([1, 1, 224, 224]) ` I am facing this error even though the target shape and...
## 🐛 Bug After creating a TCAV instance, when one calls the .interpret() method, the yielded dictionary of scores don't follow the order of the experimental sets given in input....
Hello, There is not baselines in examples in docs: https://captum.ai/api/deep_lift_shap.html but the method requires one. I tried different ones, especially ones which worked for other shap based methods in captum...
## 🚀 Feature Integrated Gradients not only for single classification task, but also for those that are based on frame-level classification, such as speech recognition, active speaker detection, etc. ##...
NoiseTunnel on LayerIntegratedGradients Perturbs Input Instead of Attribution Target (RunetimeError)
## NoiseTunnel on LayerIntegratedGradients Perturbs Input Instead of Attribution Target (RunetimeError) When applying NoiseTunnel to a LayerIntegratedGradients instance initialized on a specific model layer, the noise application happens on the...
Summary: This diff helps address the number of pyre-fixme's in the layer_feature_permutation.py file Reviewed By: jjuncho Differential Revision: D64624875
