Closed sindresorhus closed 1 year ago
I second this, if it is possible under the source licenses. I would guess it will depend on getting the main branch stable for a new release first?
Yes please, was not able to run Real-ESRGAN and lost many hours on it
@nihui makes them, ask her.
Unless someone else here is familiar with NCNN.
Just had a look at the NCNN website. It isn't clear to me that we can use it unless we completely rewrite the ML code to use the NCNN framework. Correct me if I'm wrong.
An alternative would be Docker, which I have more experience (and comfort) with. Would this be helpful? It might be one large image.
The Real-ESRGAN project provides prebuilt binaries which makes it so much easier to get started. It would be amazing if that could be provided here too.
The reality is that it's difficult to get the correct environment set up. Most threads on the internet (for example, Reddit) about Stable Diffusion or any other ML thing end up being a 100+ thread about various Python dependency issues.