J3.10.11
JRTX 4080
![Dazzastrous [4090]](https://cdn.discordapp.com/avatars/871389071686135838/c4a5f4d81d9eeb18d856295d692e2c07.webp?size=40)
Dso what are the go to settings then for Supir results
Fok
Fthen you need to do this
Fyou have SUPIR and SUPIR same folder name
Finstall like this
FR:\SUPIR_v21\SUPIR
Fdo a fresh install
Fyou can delete downloader py and move models
Ji'll try
![Dazzastrous [4090]](https://cdn.discordapp.com/avatars/871389071686135838/c4a5f4d81d9eeb18d856295d692e2c07.webp?size=16)
Fi find that CFG 6 works better with lesser hallucination
JThis worked fine. Finally I can open gradio web UI but when I try to upscale a photo I get this error on stage 1 after load both models:
JFile "D:\Prueba\STABLE DIFFUSION X\lib\site-packages\torch\nn\functional.py", line 3931, in interpolate
return torch._C._nn.upsample_nearest2d(input, output_size, scale_factors)
RuntimeError: "upsample_nearest2d_out_frame" not implemented for 'BFloat16'
return torch._C._nn.upsample_nearest2d(input, output_size, scale_factors)
RuntimeError: "upsample_nearest2d_out_frame" not implemented for 'BFloat16'
Jtorch version: 2.0.0+cu118. Do I need to upgrade it?
Fyes
Fthat is why you get error
Fuse our installer
Falso stage 1 is useless use stage 2
Jwhat version of pytorch do I need?
Fuse the auto installer 2.2.0
Jthanks
Jauto installer doesn't update my pytorch to 2.2.0, so I'm doing it manually
Fyou can delete venv
Fit should reinstall
JI also have Stable Diffusion automatic1111 installed. Too many modules has interferences with SUPIR, so I'm trying to make work all
Fthat is a big no
Fhave seperate folders
JIs on separate folders, but SUPIR load some components from stable diffusion folder
JCan't update pytorch without affect automatic1111
Fwhich components?
Jall torch components
Fthat is a very big no
Fyou will have hard time to fix
Fconflictions
JI'm near to fix it
JStage 1 now works fine. But in stage 2 I have this error: "Input type (torch.cuda.FloatTensor) and weight type (torch.FloatTensor) should be the same"
Fgood luck to fix :d
J
Jfound this on internet: "suggests that your inputs are on the GPU, but your model is not. So I’d double check whether your model is effectively on the GPU."
BHello @Furkan Gözükara SECourses, which local text-to-speech model do you recommend for longform content that is comparable to ElevenLabs (without the price tag)? Thank you
BFcocuqi is best i think
Fi plan to make ane updated tutorial with coqui hopefulyl
Dnah i think tortoise is better
Drvc
Frvc cant make text to speech as far as i know
Ftortoise was ultra slow compared to cocuqi last time i used
Dyep it can i have rvc tts
