I think it has some problem. It says music demixing is finished. But it showed only validation results (2/2). And it wait for evaluation for too long I think (4 hours).
No it’s in this repo:
Sorry link was changed by forum engine in 1st post.
I don’t have Dockerfile in my repository. I think it’s some Dockerfile used by aicrowd platform. Am I mistaken?
Hello. I wanted to make a late submission in Sound Demixing Challenge 2023:
But there is error during creating container (earlier there was no error):
Step 1/51 : FROM nvidia/cuda:10.1-cudnn7-runtime-ubuntu18.04 manifest for nvidia/cuda:10.1-cudnn7-runtime-ubuntu18.04 not found
Can you please fix it?
We also made release of demucs4 HT models trained on DnR dataset for CDX23 challenge:
Hello everyone. We made a release of our model:
And small paper about our approach:
Feel free to use it.
Final Leaderboard B the same as Final Leaderboard A
There is problem with Leaderboard C in Music Demixing Track - it sorted incorrectly.
It also will be great if “Submissions” menu will appear in “Cinematic Sound Demixing Track”.
Will it be violation of rules if I automatically fix the label noise with some opensource algorithm/model (for example UMXL)? The question arises because this opensource model can be already trained on some “non-competition” data.