This release is matching to the pre-trained model hosted on zenodo:
* __`umxhq` (default)__ trained on [MUSDB18-HQ](https://sigsep.github.io/datasets/musdb.html#uncompressed-wav) which comprises the same tracks as in MUSDB18 but un-compressed which yield in a full bandwidth of 22050 Hz.
[](https://doi.org/10.5281/zenodo.3370489)
* __`umx`__ is trained on the regular [MUSDB18](https://sigsep.github.io/datasets/musdb.html#compressed-stems) which is bandwidth limited to 16 kHz do to AAC compression. This model should be used for comparison with other (older) methods for evaluation in [SiSEC18](sisec18.unmix.app).
[](https://doi.org/10.5281/zenodo.3370486)