single modal clustering

[big-rain]

Hello! While reading your paper, I noticed that it presents clustering for unimodal cases (e.g., RNA-only, Ephys-only). However, when reviewing the open-source code, I observed that Model. forward by default expects multimodal input, and WeightedMeanFuser seems to causea significant reduction in feature values (Signal Dilution) when handling unimodal inputs (via zero-padding).

I would like to ask: When generating unimodal results, did you use either of the following approaches?

1. Manually modified the Fuser's weights during inference to allocate 100% to a specific modality?

2. Or handled missing modalities through a specific impute_check logic?

Looking forward to your reply. Thank you very much!