cog-sdxl
cog-sdxl copied to clipboard
replicate
Stable Diffusion XL training and inference as a cog model
in the dreambooth api, the model is automatically pushed without the need to create a model on replicate
### **Current Behavior** Even with the same settings & seed img2img always has significant randomness & changes in output. Occasionally output is the same as expected. ### **Problem** This behavior...
Instance can think it has loaded a fine-tune even though it hasn't - assume you have 2 fine-tunes - they are being run by sending in a url with the...
Hello @daanelson @cloneofsimo I have been trying to get the captions.csv files working but when checking the run logs on replicate.com the code just auto-captions everything. I tested the example...
See below (ignore the weird eyes): 1) original image 2) output image. For preserved areas lines are blurry.  
This issue is for stability-ai/sdxl model. When I use the model for img2img predictions and set custom width and height, the resulting output still follows the dimensions of the original...
The lora model does not load when run with the code in the following blog. blog: https://replicate.com/blog/fine-tune-sdxl When I changed the lora model loading to the load_attn_procs method as shown...
The filename required for custom captions is called "caption.csv" ([code](https://github.com/replicate/cog-sdxl/commit/a2b0c206eb9168d608e32b0991b662836dc7aedb#diff-3d114d99cf722de411abb4e7720d4e55ec1633a48d966b37c48ae692f10329feR87)) However, there is an error message that refers to it in the plural form as "captions.csv" ([code](https://github.com/replicate/cog-sdxl/commit/a2b0c206eb9168d608e32b0991b662836dc7aedb#diff-3d114d99cf722de411abb4e7720d4e55ec1633a48d966b37c48ae692f10329feR513)) It is also...
download_weights.py, download_preprocessing_weights.py not recognized as python executables, adding shebang
The captioning step is done before the segmentation step + cropping of `is face model` - meaning you will get a caption for the whole image but crop out to...
