d909b / cxplain Goto Github PK

I tried cxplain on python 3.7, but it does not work. Then, I saw some TODOs within the code saying: "TODO: remove when fully migrated in python 3.x". So I was wondering whether there was a repo with cxplain python 3.x.

I tried the CIFAR10 and MNIST example codes, and obtained the same attributions for all test images. I used the provided CIFAR10 and MNIST example codes without any modifications.
Has anyone else had the same problem?
Thanks

I try the CIFAR10 and MNIST example codes.
explainer = CXPlain(explained_model, model_builder, masking_operation, loss, num_models=5, downsample_factors=downsample_factors, flatten_for_explained_model=True) explainer.fit(x_train, y_train);
but there are the same problem when I run the codes.
ValueError: No gradients provided for any variable: ["<tf.Variable 'conv2d_130/kernel:0' shape=(3, 3, 1, 8) dtype=float32>", "<tf.Variable 'conv2d_130/bias:0' shape=(8,) dtype=float32>", "<tf.Variable 'conv2d_131/kernel:0' shape=(3, 3, 8, 8) dtype=float32>", "<tf.Variable 'conv2d_131/bias:0' shape=(8,) dtype=float32>"...
Can you please tell me how to solve this problem?
Thanks

I am getting some errors in tensorflow. Can you please tell which version of tensorflow was used while writing the notebooks?

Hello,

Is it possible to interpret results for all classes that the model has been trained on. LIME has such functionality where we can mention the class for which features should be extracted but it's too slow.

I require the cxplain feature output specific to classes that I want. Please help me with this.

Awesome library!!!