Environment Configuration Reminder

[YongJiao10]

The authors appear to have omitted the environmental details. Based on personal testing, the correct versions required to load the model (sav file) are numpy==1.19.5 and scikit-learn==0.23.2. Newer versions may result in errors.

[Ziyu20000101]

when running the nootbook meta_matching_v2.0.ipynb, some errors occur as follows:

path_v20 = os.path.join(path_repo, 'v2.0')
model_v20_path = os.path.join(path_v20, 'models')
sys.path.append(os.path.join(path_repo, "utils"))

check whether v2.0 model files exist and are up-to-date

from CBIG_model_pytorch import check_models_v20
check_models_v20(model_v20_path)

load data

from CBIG_model_pytorch import demean_norm
path_data = os.path.join(path_repo, 'data')
npz = np.load(os.path.join(path_data, 'meta_matching_example_data.npz'))
x_input = npz['x']
y_input = npz['y']
x_input = demean_norm(x_input)

print(x_input.shape, y_input.shape)

---

AssertionError Traceback (most recent call last)
Cell In[14], line 6
4 # check whether v2.0 model files exist and are up-to-date
5 from CBIG_model_pytorch import check_models_v20
----> 6 check_models_v20(model_v20_path)
7 # load data
8 from CBIG_model_pytorch import demean_norm

File ~/public/wzy_stress/meta_matching/utils/CBIG_model_pytorch.py:60, in check_models_v20(model_path)
58 # Update the MD5 hash object with the chunk
59 md5_hash.update(chunk)
---> 60 assert md5_hash.hexdigest() == MD5_ref[model_file],
61 model_file + " is not update-to-date, please download the latest version."

AssertionError: meta_matching_v2.0_model.pkl_torch is not update-to-date, please download the latest version.

it seems like i dont get a latest version of meta_matching_v2.0_model.pkl_torch, but i downloaded it according to the tips

[peterLingZijie]

@YongJiao10 Hi Yongjiao, much thanks for pointing out this. We have updated the Python packages versions (kindly note that we run on Python 3.8). Could you try again and let us know whether you could run or not? Happy research and thanks again!

[peterLingZijie]

@lily98933 Hi we recently updated the Python package versions and tested it. Could you try replicating it again and see if the error persists.
For your information, I only download and unzip meta-matching.models.v2_0.zip in the release page. My final folder structure looks like this:

Hope this helps!