about Y about xflow

[catalina17]

Hi,

For each of the samples in the train/test sets, Y contains the label and P the person from which the sample came from. They are arrays because they hold e.g. len(train) samples each for the train set.

I think you can simply create another data loader which produces tuples of (audio_vector_for_all_timesteps, image_array, label, person) for each sample, or corresponding arrays across all samples. You might need to update the models or use others, since these exploit temporal correlations across modalities.