captum

captum copied to clipboard

Zero attributions for BERT-based sentence similarity model

1

I have a BERT-based sentence similarity model, namely [bert-base-nli-mean-tokens](https://huggingface.co/sentence-transformers/bert-base-nli-mean-tokens) which takes a pair and outputs a cosine similarity score which is an indication of their semantic similarity. My goal is...

anferico

how to use layergradcam for a multi-label application?

2

I'm working with a multi-label image dataset. My inputs have the following shape: ``torch.Size([3, 224, 224])``; and my targets are all 1x33 tensors one-hot encoded, as in the following example:...

amandalucasp

Support for Identity layer in LRP

2

## 🐛 Bug Hi! Thank you for the great tool! It appears that LRP fails if the model includes `Identity` layer (https://pytorch.org/docs/stable/generated/torch.nn.Identity.html): ```python File "???.py", line 523, in my_fn attrs...

soupault

Hello everyone! I am training to put LRP into training process and involve the result from lrp into loss. I get error when I call `loss.backward()` I got this error:...

cikivan

Is there a lrp rule implemented in captum for matrix multiplication, as suggested in this document (https://openaccess.thecvf.com/content/CVPR2021/supplemental/Chefer_Transformer_Interpretability_Beyond_CVPR_2021_supplemental.pdf)? It is shown in this document that matrix multiplication breaks the relevance conservation...

oldcricket

Understand the captum output for bert

2

Hello I am new to Captum and also Bert, I have Bert model that is take as input sequence of letters of max length 2170 and output sequence of the...

SaidaSaad

TCAV for multiple instance learning

1

Hi, many thanks for this great work! I would like to ask does it make sense to use TCAV to test an attention based multiple instance network? If yes, the...

qinghezeng

AssertionError: Elements to be reduced can only be either Tensors or tuples containing Tensors.

3

when I follow the quick start, using pytorch with a two layer model(linear-relu-linear) In this sentence: cond_vals = cond.attribute(test_input_tensor, target=1) the error occured. I have checked the type of test_input_tensor----tensor,so...

soloccc

VisualizationDataRecord with a regression task?

5

I'm fine-tuning BERT (using the Transformers library) to perform a regression task and I can successfully extract the attributions. However, how am I supposed to visualize the importance of different...

RylanSchaeffer

Is there a tutorial or some documentation for LIME for models trained with tabular data?

1

## 📚 Documentation I am not able to find any documentation on using LIME for models trained on tabular data. Can you provide some?

y12uc231