Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Cannot concatenate gradients in IntegratedGradients.explain

See original GitHub issue

In the example of integrated_gradients_imdb.ipynb, the function of

explanation = ig.explain(x_test_sample, baselines=None, target=predictions) Threw the error when I tried to replicate the results.

error:

I did not modify the model and it exactly followed the example but met this error. Please take a look.

Issue Analytics

State:
Created 3 years ago
Comments:6 (3 by maintainers)

Top GitHub Comments

1reaction

Realvincentyuancommented, Jul 23, 2020

@Realvincentyuan is it possible for you to use the development version (pip install git+https://github.com/SeldonIO/alibi.git ) directly?

yup, I chose to clone the repo and import it. And it works, it’s great work, very intuitive to users. Thank you!

0reactions

jklaisecommented, Jul 22, 2020

@Realvincentyuan is it possible for you to use the development version (pip install git+https://github.com/SeldonIO/alibi.git ) directly?

Top Results From Across the Web

Integrated gradients | TensorFlow Core

This tutorial demonstrates how to implement Integrated Gradients (IG), an Explainable AI technique introduced in the paper Axiomatic Attribution for Deep ...

Integrated Gradient For Intent Classification & Ner #510 - GitHub

Hi everyone, I'm trying to use Camptum with OneNet using Allennlp, this is the model structure OneNet( (text_field_embedder): ...

Limitations of Integrated Gradients for Feature Attribution

Integrated gradients is a feature attribution method with several attractive properties, which is well suited for neural networks.

Explaining Machine Learning Models - Stanford CS Theory

Summary: Integrated Gradients is a technique for attributing a deep network's prediction to its input features. It is very easy to apply, widely...

Why did the network make this prediction?

The Attribution Problem ... Every explanation involves an implicit or explicit counterfactual ... Our attribution method: Integrated Gradients.