request Checkpoits & config
Hello,
Could you please provide the matching .ckpt and .yaml files for your BSRoformer model?
I’m getting missing keys errors when loading weights, so I think the config and weights don’t match.
Thanks in advance!
Of course, but I have to declare that I am just a porter and this is not the model I made.
This repository is only used to store and download the models that my Github project relies on.
I have now uploaded the YAML file.
Thank you. But can you upload the model in the .ckpt format? UVR accepts only this type of model.
Sorry, the model I found does not have. ckpt format. If you need to use it, you can go to my project https://github.com/CHEN-Technology/2TO5.1-OR-2TO7.1/tree/beta The logic_fsroformer folder in the Beta branch contains the code written by the original author for separating audio. After putting the model back into the logic_fsroformer/models/folder, run 1.开始分离（GPU）.bat to use GPU for separation.
Dependency can be found in requirements. txt, but due to China's GFW, I have added mirror sites within China, and you may need to remove them. There are also some dependencies required for my secondary development projects, such as pydub and ffmpeg. Triton and sageattention are dependencies I added, which seem to accelerate separation, but my testing doesn't seem to have much effect.
Finally, my secondary development project is a tool that separates stereo audio using these models and remixes it into 5.1 or 7.1 channels. If you are interested, you can give it a try.😁
Brother, how can I install it?
Brother, how can I install it?
What? Model or program?
The brother program, I find it very complete to install, don't you have any video tutorials?
The brother program, I find it very complete to install, don't you have any video tutorials?
Sorry, there are no video tutorials available, but Github provides a 7Z compressed file with an environment. Simply unzip it and run start.bat. Of course, Chinese may require translation, but the options are concise and easy to understand.