Generation doesn't work
vdyma opened this issue · 6 comments
I couldn't generate anything on the website with my prompts (Migos feat Gucci Mane drill, American platinum certified trap). It worked with a standard Eminem aggressive rap prompt, but it doesn't work with other standard prompts as well (e.g. post-teen pop talent show winner). I've waited for 30 minutes and got no result.
Hey @vdyma, we don't have enough GPUs running to serve everyone right now, we're on it. If you have your own GPU you can clone https://github.com/hmartiro/riffusion-app and https://github.com/hmartiro/riffusion-inference to run it yourself.
@hmartiro awesome work on this, I see the model is on HF: https://huggingface.co/riffusion/riffusion-model-v1, you can also setup a gradio demo for it on HF with a gpu assigned, this can scale and people can duplicate there own spaces and assign gpus based on there needs, more info on this here: https://huggingface.co/docs/hub/spaces-gpus and here: https://huggingface.co/docs/hub/spaces-sdks-gradio
@hmartiro What hardware requirements?
To have real time playback with our default settings you need to be able to generate 40 steps at 512x512 in under 5 seconds. We're using A10Gs currently in the GPU farm and it can do it in <3 seconds including the conversion to audio.
3090 is capable of realtime playback here.
you need to be able to generate 40 steps at 512x512 in under 5 seconds.
Could you place this in the readme? The "large GPU"/five seconds was a little vague for me.
Cool project, thanks!