Need help with Serverless Dreamshaper XL Worker Unstable/Throttle Issue.
Hey! I’m building a commercial DreamShaper XL image generator and need a RELIABLE serverless container on RunPod (must be able to handle high volume, with no unhealthy workers or crashing for at least several days).
Can anyone recommend a specific public container/image name (with version/tag) for DreamShaper XL serverless that’s actually working for you?
If you’ve found a stable setup, please share your image, config, and (optionally) GPU/region. Happy to pay it forward with feedback/results! 🙏
26 Replies
Unknown User•5mo ago
Message Not Public
Sign In & Join Server To View
So I did an SDXL and have had no issues but with Dreamshaper XL I keep having Unstable and throttled workers and can't figure out why
What I'm trying to do is setup like 2-3 serverless pods to call when needed I have SDXL and it seems to work fine but Dreamshaper sometimes will be fine and randomly is not.
For example I used this: leveragedpixels/dreamshaper-api built off of https://huggingface.co/imagepipeline/DreamShaper-XL1.0 (last time I tried using imagepipeline it wouldn't work so I dl'ed my down and then uplodated it to Docker Hub.
Unknown User•5mo ago
Message Not Public
Sign In & Join Server To View
So for example at this moment all if fine but randomly maybe 12 hours later all works have an unstable status.
Unknown User•5mo ago
Message Not Public
Sign In & Join Server To View
Yeah randomly I have read and AI said it's likely to Dreamshaper XL being too large or something? Wasn'ts sure if there was a way to setup my serverless to handl that.
Unknown User•5mo ago
Message Not Public
Sign In & Join Server To View
@LeveragedPixel
Escalated To Zendesk
The thread has been escalated to Zendesk!
Thank you... for full transparency I'm new to all of this and using ChatGPT to help me delpoy all of this and it suggested i post here for help because of the issues Iv'e face randomly with this serverless pod.
Do you happen to know of a good "working" dreamshaper" image/container I can use just in case mine is the issue?
I don't have any logs as I was advise if a worker goes unstable to delete and let it spin up a new one 😭
Unknown User•5mo ago
Message Not Public
Sign In & Join Server To View
Yeah my repo I used is here: https://huggingface.co/leveragedpixel/DreamShaper-XL1.0
But honestly this is a crazy project of mine. Trying to create a personal Web UI for a few models and the one issue I ran into multiple times was repos getting throttled on workers or not working on deployement so i deployed my own with the help of ChatGPT 😄
Unknown User•5mo ago
Message Not Public
Sign In & Join Server To View
I think in the end I used this perhaps not sure how to validate: https://hub.docker.com/repository/docker/leveragedpixels/dreamshaper-api/general
Because I was having container issues.
This is the issue with having 0 idea what i'm doing and having Chat GPT tell me what to do I'm sorry if I'm wasting your time.
Unknown User•5mo ago
Message Not Public
Sign In & Join Server To View
So it's a full website hosted on GitHub what code exaclty would you need I think the repo is private and I was using APIs to call.
Once again I feel like I bit off way more than I can chew.
Unknown User•5mo ago
Message Not Public
Sign In & Join Server To View
Yeah at this time I don't have any failed requests to share and I think my front and back end is private on Github.
Yeah I was like let's see if ChatGPT can help me build my own version of Midjourney or Leonardo.ai and here i am so I apologize i've learned so much and love runpod but def way over my head so far.
We can close the thread and what I'll do is create another once I get back to the failed state of having nothing but unhealthy workers and failed jobs.
Unknown User•5mo ago
Message Not Public
Sign In & Join Server To View
Oh yeah 100% but my idea is I want to create a tool a bit more available to your average joe and cheaper, but we'll see it's a crazy idea.
Thank you.
Unknown User•5mo ago
Message Not Public
Sign In & Join Server To View
I have a similar issue with my serverless endpoint for the ComfyUI API and Big Lust. Same thing: it works, and then out of nowhere, my workers get throttled. I’ve checked the logs and found no errors. Also, sometimes some workers start running by themselves, don’t generate any image, and just keep running. I have to manually shut them down to stop the process.
Unknown User•5mo ago
Message Not Public
Sign In & Join Server To View
You probably need to expand out of 4090s, + potentially add more max workers.
What is probably happening is that RunPod is essentially trying to have GPUs on standby (but if someone is scaling like crazy they can eat up all the 4090s, some businesses might have a huge influx of jobs).
So by adding more GPU types you might be able to avoid this + allow RunPod in their algorithm to swap GPUs will go down your list of priorities to swap bad workers in if they are in a state for a period of time.
Just something I've experienced before too
I did last week and I am still waiting for the response