cog-stable-diffusion
cog-stable-diffusion copied to clipboard
replicate
Diffusers Stable Diffusion as a Cog model
Cog says it's not sure about the compatibility up front, then (after a lot of downloads) it has Docker say: "Docker is missing required device driver". I figured this is...
This is just a PR to see a diff of the within-container cold start work we've done
I used ` cuda: "11.6.2"` and it seems to work. https://github.com/replicate/cog-stable-diffusion/blob/7a21fcd207cebbbf9e149b60cf9e9faa48c479ab/cog.yaml#L3
https://github.com/replicate/cog-stable-diffusion/blob/4b5af4a42cd31e873da73184df8e245602242d4a/predict.py#L26
@lleopard1704 ➜ /workspaces/cog-stable-diffusion (main) $ cog run script/download-weights Building Docker image from environment in cog.yaml... [+] Building 80.7s (16/17) => [internal] load build definition from Dockerfile 0.2s => => transferring...
Resolves #75 Currently, predictions generate images in PNG format. This PR adds a new `format` input that allows predictions to generate JPEG and WebP formats instead. If unspecified, the PNG...
``` cog build -t realms-adventurers-v3 --debug Setting CuDNN to version 11.6 Building Docker image from environment in cog.yaml as realms-adventurers-v3... $ docker build --file - --build-arg BUILDKIT_INLINE_CACHE=1 --tag realms-adventurers-v3 --progress...
WSL2 ``` $ cog run script/download-weights Building Docker image from environment in cog.yaml... [+] Building 14.0s (15/16) => [internal] load build definition from Dockerfile 0.0s => => transferring dockerfile: 2.01kB...
768x768 is the current default and the outputs tend to be pretty weird. If you're generating humans or animals, they tend to have extra heads or limbs. I know @anotherjesse...
