Sean Sube
Sean Sube
It's difficult to precisely draw in the mask canvas with a low brush strength, and the circles tend to overlap. Smooth out the strokes by reducing the number of points...
Include the git commit and tag/package version in the logs when the server starts up. - [ ] include version in bundle logs - [ ] package version - [...
https://github.com/damian0815/compel
The diagnostic scripts, for checking the runtime environment and checking model contents, are not set up to work with the Windows bundle and are not included. - [ ] figure...
Some models that have been converted by the v0.10 bundle are not working correctly when mixed with LoRAs. According to the logs, many of the necessary nodes/keys are missing, so...
Running the full ORT optimization script breaks LoRA blending: https://github.com/microsoft/onnxruntime/tree/main/onnxruntime/python/tools/transformers/models/stable_diffusion#optimize-onnx-pipeline Some nodes are removed or have their names changed: ``` [2023-03-19 16:37:11,194] WARNING: 281092 139662040862720 onnx_web.convert.diffusion.lora: could not find any...
As described in https://huggingface.co/blog/controlnet#combining-multiple-conditionings
The regular `flask` server in a `venv` works correctly with AMD hardware acceleration, but to use that in the container, the device needs to be mounted - at least. Figure...