Nicolas Fouqué

@AutoUpdatingBSoD I'm not familiar with Anaconda containers, how can they make this code work with implementations in tensorflow 2 ?

Indeed, I cannot manage to create a sequential model, currently I get ``` TypeError: The added layer must be an instance of class Layer. Found: ``` But a few weeks...

Yes I have, I usually prefer that too, but the person who showed me this package had had trouble with the functional model and managed to do it with the...

They managed to run it but only with tf 1.15 and keras 2.1.5 (and python 3.7), not tf 2+

> Sorry, I wasn't clear. What I meant was what does the statement look like where you're actually adding the model? > > I.E. The part where the program actually...

Yes yes, this one is the functional one, I thought that was what you wanted to see my sequential version is this : ``` model = Sequential() model.add(Input(shape=(seconddim, thirddim), dtype='float32',...

Ah I finally found what made it almost work the first time, the secret was in the imports This package is built by importing "pure" keras functions (i.e. not from...

If you showed what's in f it could be clearer. In my case I use a transformers pipeline as is shown in [this example](https://shap.readthedocs.io/en/latest/example_notebooks/text_examples/sentiment_analysis/Emotion%20classification%20multiclass%20example.html) Since the tokenizer is inside the...

I think this might be the same kind of discrepancy described in #2425 although we have been having it with HuggingFace transformers. Sadly we haven't had a solution for that...

If I understand the package correctly, the plots only show what the shap_values already contain, you can easily verify the outputs by executing `shap_values[0].values.sum(axis=0)+shap_values[0].base_values`, and check if the values are...