Multi-Task Automatic Synthesizer Programming Preset Dataset

<p>This data set contains different training, test, and validation data used for training the multi-task automatic synthesizer programming models in the 2023 ICASSP paper &quot;Exploring Approaches to Multi-Task Automatic Synthesizer Programming&quot;. Data is stored in .npy (numpy pickled data) format. Files are organized with &lt;split&gt;_&lt;type&gt;.npy, where &lt;split&gt; is the data split (train, test, or valid) and &lt;type&gt; is the type of data:</p> <ul> <li>serum_params are parameters for Serum data (these are already one hot encoded)</li> <li>serum_mask are masks for Serum&nbsp;parameters&nbsp;</li> <li>diva_params are parameters for Diva data (these are already one hot encoded)</li> <li>diva_mask are masks for Diva parameters&nbsp;</li> <li>tyrell_params are parameters for TyrellN6 data (these are already one hot encoded)</li> <li>tyrell_mask are masks for TyrellN6&nbsp;parameters&nbsp;</li> <li>hpss are labels that indicate the harmonic percentage of this preset 20 means 0-20%, 40 means 20-40%, 60 means 40-60%, 80 means 60-80%, and 100 means 80-100%,</li> <li>mask_single are mask for the single decoder model</li> <li>mels are mel spectrograms</li> <li>name are names of the preset</li> <li>params are the parameters of the preset</li> <li>params_single are parameters for the single decoder model</li> <li>synth are strings of which synthesizer is used</li> </ul>