Configuration error
zbruce8186 opened this issue · 7 comments
How to solve it?
Hi, the error in your screenshot will not affect the use, so don't worry. If there are no network issues, the subsequent process can continue.
set PYTORCH_CUDA_ALLOC_CONF=max_split_size_mb:?
After setting according to the prompt, it has changed again, and the setting is repeated, how should it be set?
From your log, it seems that PYTORCH_CUDA_ALLOC_CONF may not be useful.
Erase model
If you are using an erasure model, such as lama, it is recommended to set the device to cpu
Diffusion model
If you are using a diffusion model, such as runwayml/stable-diffusion-inpainting, the memory usage is partly related to the parameters of the service startup. You can reduce memory consumption by enabling the following two parameters:
On the other hand, the size of the resolution of the images being processed also affects memory usage. The higher the resolution of the image, the higher the memory usage. You can use cropper to inpaint a small area of the image. Here is an example of how to do this, you can refer to: https://www.iopaint.com/models#cropper
Many options in the recommended model need to be downloaded from the Internet, Baidu Web disk has been downloaded, but some models from the Internet download prompt connection timeout, how to solve, thank you.
Sorry for the late response, you can download the model files and configuration files manually from here. I have recorded a video demonstrating how to download one of the model files:
Area.mp4
The final directory structure looks like this:
If the link above cannot be downloaded, you can also try downloading from here: https://hf-mirror.com/runwayml/stable-diffusion-inpainting
怎样才能正确安装realesrgan？
