diff --git a/README.md b/README.md index e23e9af..22e5e0e 100644 --- a/README.md +++ b/README.md @@ -145,6 +145,7 @@ music = musiclm(['the crystalline sounds of the piano in a ballroom']) # torch.T - [x] give dynamic positional bias to self attention in AST - [ ] add a version of mulan to open clip +- [ ] support variable lengthed audio with masking in audio transformer, then implement MusicLM generating multiple samples and selecting top match with MuLaN - [ ] set all the proper spectrogram hyperparameters ## Citations