Does Runpod have an alternative to Ashley Kleynhans' github repository for creating a1111 worker?
3 days ago, I created a serverless instance by using https://github.com/ashleykleynhans/runpod-worker-a1111.
It's not accessible anymore. The https://github.com/runpod-workers/worker-a1111 doesn't seem to have the same functionality.
I would want to choose my own endpoint, loras, Ad detailer and an upscaler.
What solution do we have now to replace Ashley's work?
Does somebody as forked his repo?
GitHub
GitHub - runpod-workers/worker-a1111: Automatic1111 serverless worker.
Automatic1111 serverless worker. . Contribute to runpod-workers/worker-a1111 development by creating an account on GitHub.
31 Replies
GitHub
GitHub - ashleykleynhans/worker-a1111: Automatic1111 serverless wor...
Automatic1111 serverless worker. . Contribute to ashleykleynhans/worker-a1111 development by creating an account on GitHub.
check that one
Thank you. How can I use ad detailer, loras and upscaler with this one?
I'm not sure if it's the same tho, I don't know haven't checked it out
No it's not the same, it's the same as runpod's template https://github.com/runpod-workers/worker-a1111
Do you know if Ashley will reopen his repo?
GitHub
GitHub - runpod-workers/worker-a1111: Automatic1111 serverless worker.
Automatic1111 serverless worker. . Contribute to runpod-workers/worker-a1111 development by creating an account on GitHub.
No i dont
Does anybody has forked Ashley's repo?
We are working on getting a similar style repo live soon. Will share.
Bump
@CR-TZM we have the repo, will make sure to put this into dockerhub + provide a template tomorrow!
@GenerativeLabs looking forward to your contribution as well!
Cool, curious if it actually latest copy of Ashleys?
it should be, yes!
but we will not provide support for this yet. This is still on-going to find out how we handle this situation
so if you want to participate and invest some time into maintaining this worker, then we would be super happy about it
Yes, was planning on hosting / maintaining / supporting a version of it... want to chat or DM to figure out the best way to manage that?
Yes please, lets do it!
Great thank you a lot, I found one copy of Ashley but IDk if it's the last one but at least there is already Ad detailer and Controlnet.
GitHub
GitHub - meyiapir/runpod-worker-a1111
Contribute to meyiapir/runpod-worker-a1111 development by creating an account on GitHub.
Looks like a clone of the worker from Ashley, but I'm not sure how up-to-date this is. But if this is working for your use case, then I'm happy! We will still make sure to come up with a solution
@CR-TZM This one is up to date: https://github.com/generativelabs/runpod-worker-a1111
GitHub
GitHub - generativelabs/runpod-worker-a1111
Contribute to generativelabs/runpod-worker-a1111 development by creating an account on GitHub.
Thank you, yes I need to use SD with some specific checkpoints and Loras. I'm also using ad detailer. Do you know if it's possible to use controlnet openpose in txt2img?
If yes, should I just use the img2img JSON for controlnet and add it at the end of my txt2img JSON?
Yes you can use ceontrolnet with txt2img. No, don't use img2img json -- use txt2img, then you pass the base64 encoded image as value to the input_image arg in the controlnet itself
Ok I'll test it thanks
{
"input": {
"api": {
"method": "POST",
"endpoint": "/sdapi/v1/txt2img"
},
"payload": {
"override_settings": {
"sd_model_checkpoint": "cyberrealistic_v42",
"sd_vae": "vae-ft-mse-840000-ema-pruned.safetensors"
},
"override_settings_restore_afterwards": true,
"refiner_checkpoint": "",
"refiner_switch_at": 0.8,
"prompt": "my prompt",
"negative_prompt": "",
"seed": -1,
"batch_size": 1,
"steps": 25,
"cfg_scale": 7,
"width": 512,
"height": 720,
"sampler_name": "DPM++ SDE",
"sampler_index": "DPM++ SDE",
"scheduler": "karras",
"restore_faces": false,
"alwayson_scripts": {
"controlnet": {
"args": [
{
"enabled": true,
"module": "openpose",
"model": "control_v11p_sd15_openpose",
"weight": 1.0,
"input_image": base64 image,
"resize_mode": "Crop and Resize",
"lowvram": false,
"processor_res": 64,
"threshold_a": 64,
"threshold_b": 64,
"guidance_start": 0.0,
"guidance_end": 1.0,
"control_mode": "Balanced",
"pixel_perfect": false
}
]
}
}
}
}
}
I used this code with image and input_image, and both succeeded in generating but none of them used controlnet openpose
This is the image I used as input
Hello @GenerativeLabs , do you know what is wrong with my API call that openpose is not used?
whats wrong
The picture generated with this API call doesn't use the pose from the picture above.
check the logs maybe theres an error there ( im not sure whats wrong )
Are you sure the controlnet models are there? Can call GET /controlnet/model_list to verify + confirm
Also separately noticed your processor_res is set very low at 64. This can actually cause poor detail recognition in your input image. Try increasing it to at least 512. You could even go higher, like 768 or 1024, if you can spare the processing time and memory usage.
Thank you for the help, I probably didn't have the openpose models. I wanted to try again but Ashley deleted his docker image. So I recreated the network volume and docker image by following step by step the whole tuto.
Now I have this error when I try to make an API call:
2024-08-16T15:44:25.039554668Z /usr/bin/env: 'bash\r': No such file or directory
This error comes from my docker image, it's probably because I created it on windows. Idk. Do you have a working docker image so I can test to use openpose now that I added the model?
I don't have a working docker image at the moment that's current. I'm working on putting something together though, just can't promise exactly when that'll be available. As for the error, one way that often debug errors like this (No such file or directory) is by increasing the idle timeout, hitting the endpoint, then connecting to the server using web terminal and seeing what's up with pathing by poking around the filesystem.
Ok no problem I'll try to solve this issue. Thank you
I found the solution for the bash error.
It's caused by my Windows env. To solve it you need to add one line in the Dockerfile :
#- Add RunPod Handler and Docker container start script
COPY start.sh rp_handler.py ./
COPY schemas /schemas
#- Fix line endings in the start.sh script (this is the line added)
RUN sed -i 's/\r$//' /start.sh
#- Start the container
RUN chmod +x /start.sh
ENTRYPOINT /start.sh
Concerning the openpose model I'm still trying to make it work
I think I found the issue for openpose too. It's due to the input image 🥲
The openart image that I use doesn't get analyzed by the preprocessor. I used another picture and now it's working well
@CR-TZM that's great to hear!