Jarredou

It could be because of problem with envs using latest cuda versions (>=12.x), which leads to any onnx models processing done on CPU instead of GPU (and so being extremely...

@HyperK3122 Latest Colab fork version is available here: https://github.com/jarredou/MVSEP-MDX23-Colab_v2/

Band splitting and keep only demucs_ft above 15khz for vocals ?

I think there's is an issue when the phase cancellation occurs between mixture and separated stems. As some models have a frequency cutoff, when merging the mixture (fullband) and the...

@turian I'm mostly using my fork of this script (https://github.com/jarredou/MVSEP-MDX23-Colab_v2/), which, to separate vocals, uses mostly InstVocHQ model (which is available in UVR) and a bit of the VitLarge one...

Please report any bug or request related to the colab fork here by now: https://github.com/jarredou/MVSEP-MDX23-Colab_v2/issues And to answer your question, it is not planned at the moment. I have some...

@alexclarke236 Would you like to share the checkpoints you've trained ? Best way is to host them on a file-sharing site and post the link here like previous users have...

Description: MDX23C Drums elements separation model (to apply on drums-only audio) n_fft = 2048 instead of default 8192 was used for more lightweigted required resources. Baseline training (141 epochs) was...

I've started working on a more "resume-friendly" fork a while ago with the --resume CLI args, and saving optimizer, scheduler states + epoch, best_sdr and last training loss values within...

It would require more work for a PR, like I said it's not bulletproof in its current state and can lead to some errors, but since few months, I don't...