zennit-crp
zennit-crp copied to clipboard
rachtibat
An eXplainable AI toolkit with Concept Relevance Propagation and Relevance Maximization
## What does this PR do? It adds a custom `Dataloader` inheriting from `torch.utils.DataLoader` along with a custom `sampler`. These are dynamics in meaning that they'll retrieve indices only known...
Hi @rachtibat, the `run_distributed` method of the `FeatureVisualization` class does not take into account the actual `batch_size` for the multi-target case. Maybe include something like: ``` if n_samples > batch_size:...
We know that activation != relevance. To localize concepts in input space, we initialize CRP with channel activations using the start_layer argument of the CondAttribution class. This works well for...
A simple solution for the current problem that torch.take is not implemented for MPS. 
### Issue: Common Import Problem with CRP and Zennit 0.4.6 #### Background The CRP package currently enforces the use of `zennit==0.4.6`, while the latest version of [zennit](https://github.com/chr5tphr/zennit) is 0.5.1. Users...
