Compatibility Issue with TensorFlow 1.15.5 and SHAP Library: Seeking Guidance

Question

I am writing to seek assistance with a compatibility issue I am facing while using TensorFlow 1.15.5 in conjunction with the SHAP library.

Here is a portion of the code where the problem occurs, as it is a very long script:

type e = shap.DeepExplainer(([self.X1, self.X2], self.H),
                               [self.training_data1, self.training_data2],
                               session=self.sess, learning_phase_flags=[self.is_train])
        shap_values = e.shap_values([self.training_data1, self.training_data2])

The specific error message I am encountering is:

File "C:\Users\Python37\lib\site-packages\shap\explainers\deep\deep_tf.py", line 290, in run
for t in self.learning_phase_flags:
AttributeError: 'TFDeepExplainer' object has no attribute 'learning_phase_flags'

I've tried different versions of the SHAP library, including some unfinished versions, in an attempt to find a solution. However, I have not been able to resolve the issue. Additionally, extensive searches on Google did not yield any viable solutions.

I suspect the problem may be related to compatibility issues between TensorFlow 1.15.5 and SHAP. Given the importance of this code for a project I'm working on, any guidance or insights you could provide would be immensely valuable.

Expected Outcome:

I expected the code to run without any errors, allowing me to proceed with my project involving the integration of TensorFlow and SHAP.

Actual Outcome:

However, I continue to face the same error related to the learning_phase_flags attribute, which is preventing me from moving forward.