replicate experiments on the LLama model.

[xyzCS]

Hello, thank you for sharing your code. Now I want to replicate your experiments on the LLama model, mainly to experiment on the compound_words dataset. However, the data you provided should be based on the Pythia model, so I need to rebuild it for the LLama dataset. I noticed that there is a file called make_feature_datasets.py, but when I tried to run this file to build the data, I found that the file "PILE_TEST_PATH = '/home/gridsan/groups/maia_mechint/datasets/pile-test.hf'" is missing. Can you please provide this file? Or if I need to replicate your experiments on LLama, do you have any suggestions?

[xyzCS]

Another problem is that when I try to run probing experiments on the compound_words dataset for Pythia, I get the following error message: FileNotFoundError: [Errno 2] No such file or directory: 'sparse-probing-paper/results/activation_datasets/pythia-70m/compound_words.pyth.24.-1.next_token/mlp.hook_post,none/0.all.pt' on line 82 of the probing_experiment.py file, where the author does not provide activation_probing_dataset for different layers.