Question about duplicate TUEV label

[Aceticia]

Hi all thanks for the great work. I have a question about how you are forming the label for temple events corpus. In your script making the data, every line in the .rec file creates a sample and a label. From what I see, there are two doubts I have about this treatment:

1. In many occasions, multiple sensors have the same label simultaneously. This would introduce duplicate samples into the dataset.
2. There could be overlapping events occasionally between different sensors. Your treatment would introduce two different samples with different labels into the dataset.

When processing, you do store the offending channel id, which would disambiguate these two issues, but TUEVLoader does not use it:

    def __getitem__(self, index):
        sample = pickle.load(open(os.path.join(self.root, self.files[index]), "rb"))
        X = sample["signal"]
        if self.sampling_rate != self.default_rate:
            X = resample(X, 5 * self.sampling_rate, axis=-1)
        Y = int(sample["label"][0] - 1)
        X = torch.FloatTensor(X)
        return X, Y

Did you perhaps fix this somewhere else or in some unpublished code? Maybe I missed some preprocessing steps? Please correct me if I misread something since I'm on 3 hours of sleep here :)