5fa1a76
1
The CLAP model uses a SWINTransformer to get audio features from a log-Mel spectrogram input, and a RoBERTa model to get text features.
