after many hours of trying im still missing something. i got a method that works in conda but not working in regular python project with packages installed. same model of course. def checkIfRowValid(model: Model): int_list = np.random.choice([0, 1], size=(1, 90)) # works in conda y_hat = model.predict(int_list, batch_size=1) # works in conda is says: 'SymbolicTensor' object has no attribute 'numpy' I saw numpy doesnt work with without eager mode but why does it work in conda? How can i make it work in regular python without eager mode also ? because eager mode is really slow . Continue reading...