Rob Ballantyne

This may be because I didn't include a section to enable GPU. A fairly big oversight that I'll rectify soon. Could you try appending the following to the compose file...

Thanks for the update. The ComfyUI process is ending too quickly. I need to change the supervisor startsecs to 5 (build/COPY_ROOT/etc/supervisor/supervisord/conf.d/comfyui.conf) to ensure its properly marked as a failure and...

What is your build environment? I have completed a successful local build prior to pushing changes on an intel i5, 4gb ram

I have no idea, unfortunately. To debug you can set +x on the build scripts to find exactly where the failure occurs. I only build on Ubuntu (same as the...

I have re-worked some of the base images and an update is imminent. Hopefully it will address these issues as there is a new docker-compose.yaml and a slightly modified build...

> I just pulled this and tried it with latest-rocm, and am experiencing the same issue > > ``` > supervisor-1 | Starting ComfyUI... > supervisor-1 | Starting ComfyUI... >...

No, this needs fixing. I use backblaze which doesn't have the same issue but I intend to properly support s3 regions.

@ultimatech-cn when you've opened the ComfyUI URL, you can manually append rp-api to load the API playground. It should look something like this: `https://xxxx-8188.proxy.runpod.net/rp-api` I'll eventually add a link to...

@Daggle24 You're absolutely right. The compose file is referencing ghcr.io/ai-dock/jupyter-pytorch:latest-cuda - This was fine until recently when I updated the PyTorch images Please try rebuilding with the base image set...

Probably, yes. I did not finish that function as I decided not to support base64 output for the time being. I'll add it back when I get chance but using...