Questions about the generation of feature vectors for the dataset

[invisprints]

Hello,

I carefully reviewed your repository code, issues, and the code related to the DSMIL data generation section. However, I could not find more detailed information about train_feats.npy in TCGA_LungCancer_CLIPFeat.

TOP/Datasets_loader/dataset_TCGA_LungCancer.py

Line 199 in 8356b0d

self.patch_corresponding_slide_label = np.load(os.path.join(feat_dir, "train_corresponding_slide_label.npy"))

I would like to ask how the four important files in your dataset — train_feats.npy, train_corresponding_slide_label.npy, train_corresponding_slide_index.npy, and train_corresponding_slide_name.npy — were generated.