Captured handler exception
while using runpod sdxl api i usually get this in between running code, when i call the output, and it gets fixed after some time ,
p.s there is no chnage in code (because sometimes for the same payload it works sometimes it shows this)
{
"error": "'NoneType' object has no attribute 'split'"
}
In the logs i see this
"Captured Handler Exception"
50 Replies
@CodeBroker is also reporting issues with this API, seems its broken. @Merrell @flash-singh
Interesting part is, when I try to test on your website with the same parameters, it does work. But when it comes to a POST it does not
Are you using the RunPod one or did you use quick deploy?
I used quick deploy
But as API reference I use this page https://doc.runpod.io/reference/stable-diffusion-xl
Do you see any errors in your logs?
No only "COMPLETED"
I assume @minu is also using quick deploy otherwise they would have no access to the logs.
There should be something in the logs
I will make a new request
give me couple seconds
You can filter by error log level like this
okay
yes I have one
Any others?
Only this
others are info logs
Well there is a problem right here:
https://github.com/runpod-workers/worker-sdxl/blob/main/src/rp_handler.py#L115
GitHub
worker-sdxl/src/rp_handler.py at main · runpod-workers/worker-sdxl
RunPod worker for Stable Diffusion XL. Contribute to runpod-workers/worker-sdxl development by creating an account on GitHub.
Its sending a list instead of a string which is not supported.
Should only affect validation of input though.
yes am using quick deploy
In the end shouldn't I supposed to send a Json file?
I will send the same request on your page with the same format
Not my page, I don't work for RunPod, I am just a community member 😁
Alright now I get the same error, but in order to understand the request type could you give an example for me.
This was my request;
{
"input": {
"init_image": null,
"mask": null,
"prompt": "Positive Prompt: Magician, clash of clans style",
"negative_prompt": "",
"height": 512,
"width": 512,
"num_inference_steps": 50,
"guidance_scale": 7.5,
"num_outputs": 1,
"seed": 963652645
}
}
Oh really
Sorry man
That height and width are not suitable for SDXL, SDXL should be 1024x1024. SD 1.5 is 512x512.
2024-03-19T05:33:15Z error starting: Error response from daemon: No such container: 77tt3raewfey9h-1
This looks like a different kind of issue
thios was i guess also the error on the worker, at that time
i cant understand properly, when i got the error i noted the time, and i am showing errors i got for that time
it does support 512 too
Alright found the problem
{
"input": {
"prompt": "Positive Prompt: Magician, clash of clans style",
"negative_prompt": "",
"height": 1024,
"width": 1024,
"num_inference_steps": 50,
"guidance_scale": 7.5,
"num_images": 1,
"seed": 963652645
}
}
This needs to be the parameter. It has couple differences
But to keep the issue up, output is reaallyy bad, it is not possible to get the images
it does need an update and if possible input parameters should be same as the previous ones
What do you mean the input parameters should be the same as the previous ones?
I think you need to specify
s3Config for an S3 compatible bucket for it to store your images.compared to Stable Diffusion 1.5 and 2 parameters and their naming changed, which couses conditional access for every model
What do you mean by this?
I was trying to say this part, the output is really bad, you can not get any information from here
You need to base64 decode those images.
Send a request | RunPod Documentation
The method in which jobs are submitted and returned.
I have one that is not the problem
problem here is, taking the right json array in order to download these images
I would just use
image_urlfor example why do we have "image_url"
It is not array, "images" is
but they conflict each order I suppose
But makes no difference, since you are only generating 1 image
I am making multiple
2 for sdxl
Oh I see, yeah then image_url is just cluttering the output, you're right.
yep
Bro, do you know if I can attach more than one AWS bucket to the same runpod template?
Because I will need to use the same runpod template for more than one app... and each app has its own S3 bucket...
Mmmm, I dont think you can use it like this way from the rp's handler :https://docs.runpod.io/serverless/workers/development/environment-variables#python-code-for-s3-uploads
Use environment variables | RunPod Documentation
Incorporating environment variables into your Handler Functions is a key aspect of managing external resources like S3 buckets.
but i think you can make your own custom code and use S3 on your code using libraries to upload them to your preconfigured buckets ( custom code )
Yeah, I also told him the same thing
hahaha
okok i think thats the only simple way then
Maybe simple, maybe not depending on your coding proficiency, but there is always ChatGPT etc to help out 🤣
But its basically the only way, simple or not.
ah yes that made my life waay simpler
Yeah me too, I can do it myself, but why waste hours, when AI can help get it done in minutes
yea i actually got another idea: adding that custom code and integrating it into custom runpod's library
By the way @rafael21@ I don't recommend using S3. AWS egress costs are ridiculously expensive. Rather use something like Cloudflare R2 which is S3 compatible but has zero egress costs.
Lets move this chat to #AWS S3 about that i just noticed 😄
yeah not sure why @rafael21@ brought it here, this is a different topic