Model not available anymore ? Google drive Error 404

[jarredou]

The model is not available anymore, gdown link from the install script throws a 404 error. Is this done on purpose ?

[lucellent]

Same issue. Looks like it doesn't work anymore

[deton24]

@inagoy Looks like the model might be still online if you log in to this page https://huggingface.co/inagoy/drumsep/resolve/main/modelo_final.th

edit. No. Actually, everybody can actually use your own credentials (username without email) and it says repository not found, and rejects if credentials are incorrect, and it's enough to be logged in to your regular account to see repository not found error.

(Google cache: https://webcache.googleusercontent.com/search?q=cache:mRRhbeZck5QJ:https://huggingface.co/inagoy/drumsep/tree/main&cd=2&hl=pl&ct=clnk&gl=pl_)

Please reupload the modelo_final.th file here and, if you had more time, please repair the Colab as well.

That link no longer works. https://drive.google.com/uc?id=15eooSWblsJLwJ7EauxGrlmoSypgwG8ug You could exceed GDrive quota, I'd rather recommend upload model file to releases page on GH to overcome file limit and use that link in Colab instead.

[HENDRIX-ZT2]

Also interested in trying this model, hoping for a reupload. :)

[Wattbag]

Hi, as a drummer & hdemucs user, I'm totally interested by your project too !! So please @inagoy reload your model :+1: : it sounds pretty cool to unmix kick/snare/tom/cymbals indeed ! Muchos Gracias :

[deton24]

Someone else than us, uploaded it it in few parts in an issue of Demucs repo. https://github.com/facebookresearch/demucs/issues/422#issuecomment-1537473242 Later you got the full link pasted by someone.

I found it in Google like a week or more ago.

I tried to contact the Inagoy via email, to ask for reupload in this repo, but haven't received the reply, so I thought it might be better to share it at this point.

Fixed Colab https://colab.research.google.com/drive/1wws3Qm3I1HfMr-3gAyW6lYzUHXG_kuyz?usp=sharing (reworked, very simple for now, because some code didn't want to work, and got deleted, edit. GPU enabled now, overlaps/shifts tweaking added for better results)

You can also use drumsep in UVR 5 GUI. Go to UVR settings and open application directory. Find the folder "models" and go to "demucs models" then "v3_v4" Copy and paste both the .th model file and .yaml file, and it's good to go.

https://cdn.discordapp.com/attachments/708579735583588366/1124699600717094932/drumsep.yaml

[Wattbag]

Great 👍 thanks

[godzfire]

Someone else than us, uploaded it it in few parts in an issue of Demucs repo. facebookresearch/demucs#422 (comment) Later you got the full link pasted by someone.

I found it in Google like a week or more ago.

I tried to contact the Inagoy via email, to ask for reupload in this repo, but haven't received the reply, so I thought it might be better to share it at this point.

Fixed Colab https://colab.research.google.com/drive/1wws3Qm3I1HfMr-3gAyW6lYzUHXG_kuyz?usp=sharing (reworked, very simple for now, because some code didn't want to work, and got deleted, edit. GPU enabled now, overlaps/shifts tweaking added for better results)

You can also use drumsep in UVR 5 GUI. Go to UVR settings and open application directory. Find the folder "models" and go to "demucs models" then "v3_v4" Copy and paste both the .th model file and .yaml file, and it's good to go.

https://cdn.discordapp.com/attachments/708579735583588366/1124699600717094932/drumsep.yaml

I'm trying to add this to UVR and seem to have it halfway. It shows the drumsep as the demuc model, but the stems are the default ones as shown and not the proper drums. I have the items in the proper place. What am I doing wrong?

[deton24]

It's probably because the GUI wasn't written with these specific instruments in mind, hence it labels them wrong. Can't tell if editing any file under: ultimatevocalremovergui/models/Demucs_Models/model_data/ would do the trick, but that's possibly the place to look for some config editing, if you don't see a pattern in other yaml files in v3_v4 folder of what can be wrong, and what the file yaml file I posted is missing in its content to label the stems correctly. That's more an issue for https://github.com/Anjok07/ultimatevocalremovergui/tree/master repo

sob., 14 paź 2023 o 23:24 godzfire @.***> napisał(a):

Someone else than us, uploaded it it in few parts in an issue of Demucs repo. facebookresearch/demucs#422 (comment) https://github.com/facebookresearch/demucs/issues/422#issuecomment-1537473242 Later you got the full link pasted by someone.

I found it in Google like a week or more ago.

I tried to contact the Inagoy via email, to ask for reupload in this repo, but haven't received the reply, so I thought it might be better to share it at this point.

Fixed Colab https://colab.research.google.com/drive/1wws3Qm3I1HfMr-3gAyW6lYzUHXG_kuyz?usp=sharing (reworked, very simple for now, because some code didn't want to work, and got deleted, edit. GPU enabled now, overlaps/shifts tweaking added for better results)

You can also use drumsep in UVR 5 GUI. Go to UVR settings and open application directory. Find the folder "models" and go to "demucs models" then "v3_v4" Copy and paste both the .th model file and .yaml file, and it's good to go.

https://cdn.discordapp.com/attachments/708579735583588366/1124699600717094932/drumsep.yaml

I'm trying to add this to UVR and seem to have it halfway. It shows the drumsep as the demuc model, but the stems are the default ones as shown and not the proper drums. I have the items in the proper place. What am I doing wrong? [image: image] https://user-images.githubusercontent.com/2067820/275289370-1ba98359-f6e0-4b1e-9e4e-f33d5d007024.png [image: image] https://user-images.githubusercontent.com/2067820/275289380-092856eb-465a-43de-847b-ef7f1c7d7463.png

— Reply to this email directly, view it on GitHub https://github.com/inagoy/drumsep/issues/3#issuecomment-1763184956, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIJ3EHDEO42CBNDVS74O4VDX7L7KBAVCNFSM6AAAAAAWRJ3QRGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTONRTGE4DIOJVGY . You are receiving this because you commented.Message ID: @.***>

[godzfire]

It's probably because the GUI wasn't written with these specific instruments in mind, hence it labels them wrong. Can't tell if editing any file under: ultimatevocalremovergui/models/Demucs_Models/model_data/ would do the trick, but that's possibly the place to look for some config editing, if you don't see a pattern in other yaml files in v3_v4 folder of what can be wrong, and what the file yaml file I posted is missing in its content to label the stems correctly. That's more an issue for https://github.com/Anjok07/ultimatevocalremovergui/tree/master repo sob., 14 paź 2023 o 23:24 godzfire @.***>

I used your file. Does it work & show properly for you?

[jarredou]

@godzfire You can also use DrumSep model with Demucs command line, you only need Python installed and then to run the command provided here: https://github.com/facebookresearch/demucs/issues/422#issuecomment-1501932895

[deton24]

The separation itself used to work following that instruction. Haven't you tried separating yet?

niedz., 15 paź 2023, 04:07 użytkownik Jarredou @.***> napisał:

@godzfire https://github.com/godzfire You can also use DrumSep model with Demucs command line, you only need Python installed and then to run the command provided here: facebookresearch/demucs#422 (comment) https://github.com/facebookresearch/demucs/issues/422#issuecomment-1501932895

— Reply to this email directly, view it on GitHub https://github.com/inagoy/drumsep/issues/3#issuecomment-1763236702, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIJ3EHG5FZLW7OUIHJRN3T3X7NAO7AVCNFSM6AAAAAAWRJ3QRGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTONRTGIZTMNZQGI . You are receiving this because you commented.Message ID: @.***>

[godzfire]

@godzfire You can also use DrumSep model with Demucs command line, you only need Python installed and then to run the command provided here: facebookresearch/demucs#422 (comment)

@jarredou Unfortunately I don't know how to use command line and do GUI.

It's probably because the GUI wasn't written with these specific instruments in mind, hence it labels them wrong. Can't tell if editing any file under: ultimatevocalremovergui/models/Demucs_Models/model_data/ would do the trick, but that's possibly the place to look for some config editing, if you don't see a pattern in other yaml files in v3_v4 folder of what can be wrong, and what the file yaml file I posted is missing in its content to label the stems correctly.

@deton24 I compared the yaml files with other ones and there's no instrument information inside them, so it's not that. I know it's housed somewhere linked to what demuc model is selected because I auto downloaded htdemucs_6s and that one worked just fine.

[deton24]

Anyway, I understand that the problem is just with wrong labels to the stems and separation using Drumsep inside UVR 5 GUI normally works for you?

czw., 19 paź 2023, 01:30 użytkownik godzfire @.***> napisał:

@godzfire https://github.com/godzfire You can also use DrumSep model with Demucs command line, you only need Python installed and then to run the command provided here: facebookresearch/demucs#422 (comment) https://github.com/facebookresearch/demucs/issues/422#issuecomment-1501932895

@jarredou https://github.com/jarredou Unfortunately I don't know how to use command line and do GUI.

It's probably because the GUI wasn't written with these specific instruments in mind, hence it labels them wrong. Can't tell if editing any file under: ultimatevocalremovergui/models/Demucs_Models/model_data/ would do the trick, but that's possibly the place to look for some config editing, if you don't see a pattern in other yaml files in v3_v4 folder of what can be wrong, and what the file yaml file I posted is missing in its content to label the stems correctly. That's more an issue for https://github.com/Anjok07/ultimatevocalremovergui/tree/master repo sob., 14 paź 2023 o 23:24 godzfire @.***> napisał(a): @deton24 https://github.com/deton24 I compared the yaml files with other ones and there's no instrument information inside them, so it's not that. I know it's housed somewhere linked to what demuc model is selected because I auto downloaded htdemucs_6s and that one worked just fine.

[image: image] https://user-images.githubusercontent.com/2067820/276433537-a2bdf97d-2a9d-472b-8c2c-01815194fb2a.png

— Reply to this email directly, view it on GitHub https://github.com/inagoy/drumsep/issues/3#issuecomment-1769577101, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIJ3EHBN3R67VXUDFOFRDYTYABRCFAVCNFSM6AAAAAAWRJ3QRGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTONRZGU3TOMJQGE . You are receiving this because you were mentioned.Message ID: @.***>

[godzfire]

@deton24 From what I can tell, yes, the drumsep does at least seem to run okay from the tests I've done.

For the record, the label issue appears to be UVR specific: https://github.com/Anjok07/ultimatevocalremovergui/issues/915#issuecomment-1771187699

[godzfire]

If anyone wants a full copy of the modelo_final.tn file, someone uploaded it here: https://drive.google.com/file/d/1mbhHbP47fWXC4gnqAJVCezgxmh-h5uVr/view

[deton24]

New drumsep model by jarredou and Aufr33: https://docs.google.com/document/d/17fjNvJzj8ZGSer7c7OFe_CNfUKbAxEh_OBv94ZdRG5c/edit#heading=h.2u19k7ty9b00

Works with UVR. Colab attached. Better SDR than the original Drumsep. Expanded dataset and better model architecture (MDX23C).

Still, I'd recommend using it only on (e.g. separated) drums.