Tutorial "02_training_analysis.ipynb" Error: <path>/python3.10/site-packages/mvtcr/evaluation/Mtrics.py line 106

I'm not sure if I didn't install the env sucessfully, I just do as the readme said. When I use "Haniffa" data to test the tutorials "[01_preprocessing.ipynb](https://github.com/SchubertLab/mvTCR/blob/master/tutorials/01_preprocessing.ipynb)" & "02_training_analysis.ipynb"

The preprocessing works very well. But when it comes to train the model, the error happend...

It's time i have fix it, so i can give a blur description that it happened because of Metrics.py，numpy Error

at:（I have modified the scripts below, it can solve the error happened on this issue）


def get_knn_f1_within_set(latent, column_name):
    con = latent.obsp['connectivities'].todense().astype(bool)         # source [latent.obsp['connectivities'].A] > error: no attr .A
    nearest_neighbor_label = [latent.obs[column_name].values[np.array(row)[0]].tolist()[0] for row in con]    # below
    #source [nearest_neighbor_label = latent.obs[column_name].values[row].tolist()[0] ] ....
    # because row is type np.matrix, so can‘t use row[0] to get as 1-d index 
    # I transfer it to np.array then [0] to fix this issue
    labels_true = latent.obs[column_name].values
    if torch.is_tensor(nearest_neighbor_label[0]):
        nearest_neighbor_label = [str(el.item()) for el in nearest_neighbor_label]
        labels_true = [str(el.item()) for el in labels_true]
    result = classification_report(labels_true, nearest_neighbor_label, output_dict=True)
    result = result['weighted avg']['f1-score']
    return result

I have write down after ’#‘ to describe the scripts I modified

Then it can train sucessfully, with 'pseudo_metric' as params_optimization name。

Has anyone encountered such a problem? Or what could be causing this, package version or other factors? Thank you very much.

SchubertLab / mvTCR

Tutorial "02_training_analysis.ipynb" Error: <path>/python3.10/site-packages/mvtcr/evaluation/Mtrics.py line 106 #16