captum
captum copied to clipboard
pytorch
Model interpretability and understanding for PyTorch
## ❓ Questions and Help Saving images from captum is somewhat difficult. When I call `fig,ax = viz.visualize_image_attr` I get a figure and an axis in return. The problem is...
My CNN model takes input of size (batch, 1, 40000). When I try to use data with that size for the `inputs` argument of `DeepLift(model).attribute()`, I get `TypeError: chunk(): argument...
I have a single layer RNN model that takes BERT token embeddings as input. When calculating attributions using Integrated Gradients, I don't seem to get any negative attributions at all....
## 🚀 Feature Give `visualize_image_attr` a flag that allows to save the figure in a given path, might as well implement that new flag for every other visualization function ##...
## 🐛 Bug When LRP is used with a custom propagation rule, calling `attribute` more than a single time fails because the custom rule is removed in https://github.com/pytorch/captum/blob/5543b4a8d256f2fa6d43eb3ac8675235f996c9fe/captum/attr/_core/lrp.py#L382 ## To...
## ❓ Questions and Help In most docstrings of captum attribution methods, there is an example where the model is said to return class probabilities: ``` >>> # ImageClassifier takes...
Summary: Update rst file to include LayerFeaturePermutation. Differential Revision: D56580319
## 🐛 Bug Between captum 0.5.0 and 0.6.0 the arguments `grad_input` and `grad_output` in the DeepLIFT implementation switched from being a tuple of length 1 to just being a PyTorch...
CLRP uses different inital LRP score, and I want to use lrp method to compute CLRP. Is it possible?
Hi, I'm trying to do first experiments with captum, with a pretrained BertForMultipleChoice model over this [published Caseholder dataset](https://huggingface.co/datasets/casehold/casehold), and following this [BERT SQUAD tutorial](https://github.com/pytorch/captum/blob/master/tutorials/Bert_SQUAD_Interpret.ipynb). I've posted [my notebook on...
