AlxndrMlk/blogs-code

Bug in "Beyond the basics - Causal Discovery (causalpython.io).ipynb" ?

JuergenKanz opened this issue · 5 comments

JuergenKanz commented

I tried to run your notebook "Beyond the basics - Causal Discovery (causalpython.io).ipynb".
Three attribute errors occurred:

  1. true_dag = DAG.scale_free(n_nodes=10, n_edges=15, seed=18)
  2. B_und = DAG._graph_to_adjmat(G_und) in File "e:\anaconda3\lib\site-packages\castle\datasets\simulator.py:184, in DAG.scale_free(n_nodes, n_edges, weight_range, seed)"
  3. return nx.to_numpy_matrix(G) in File "e:\anaconda3\lib\site-packages\castle\datasets\simulator.py:64, in DAG._graph_to_adjmat(G)" with the message: "module 'networkx' has no attribute 'to_numpy_matrix'"

I have installed "castle version 1.0.3" and "networkx 3.0".

I hope this will help for further improvement.

AlxndrMlk commented

Hi @JuergenKanz

Thank you for opening the issue. I appreciate it!

I added the missing conda environment file and environment installation instructions.

Can you confirm if the code works for you when you follow the instructions from newly added README.md

JuergenKanz commented

Hi @AlxndrMlk

Yes, the newly added instructions are working. Thanks for support.

  • 👍1
AlxndrMlk commented

Happy to hear that @JuergenKanz

Thank you for the contribution!

JuergenKanz commented

Hi @AlxndrMlk

Please forgive me, but I dislike to work with different Python environments. Hence, I checked what I can do to work with the latest versions of all needed packages. My described problem above is related to the latest networkx = 3.0 version. As already mentioned, the attribute error "module 'networkx' has no attribute 'to_numpy_matrix'" is the issue. 'to_numpy_matrix' is replaced in networkx 3.0 with 'to_numpy_array'.
Therefore, I replaced in Line 64 and Line 563, 'to_numpy_matrix' with 'to_numpy_array' in file E:\anaconda3\Lib\site-packages\castle\datasets\simulator.py
Your notebook is now running perfectly even with the latest package versions. Please don't hesitate to share this information with your readers.

AlxndrMlk commented

Thank you for the update @JuergenKanz

I understand that you don't like to work with Python environments. Thank you for sharing your opinion and recommendations.

I am closing the issue now. If there's anything more I can do for you, feel free to let me know.