RunPod

R

RunPod

We're a community of enthusiasts, engineers, and enterprises, all sharing insights on AI, Machine Learning and GPUs!

Join

⚡|serverless

⛅|pods-clusters

serverless instances having issue with caching container image?

We are seeing an increased egress cost a repo of Artifact Registry of GCP since a few days ago. Two serverless instances uses the container image from that repo. The repo is on US and the serverless instances are in Europe. The access to the repo is done with Registry Credentials configured on Docker Configuration of serverless....

Question about Serverless V2 API Payload for Automatic1111 Inpainting

Hi, I'm trying to perform inpainting using the RunPod V2 API (/runsync) with my Serverless endpoint ID (which runs an Automatic1111-compatible image). I'm sending a JSON payload in the input object that includes prompt, init_images (as a list with one base64 string), mask (as a base64 string), denoising_strength, inpainting_fill, inpaint_full_res, and mask_blur. However, the generation ignores the init image and mask. The response info from the backend shows "is_using_inpainting_conditioning": false....

Issue with Websocket latency over serverless http proxy since runpod outage

We have a runpod serverless endpoint which we have been using to stream frames over direct one-to-one websocket. We have a lightweight version of this endpoint we've been using that streams simple diagnostic images, and a production version that streams AI generated frames. Frames are configured to stream at 18fps in both cases to create an animation. We now see that both versions of this endpoint fail to stream frames at a reasonable rate, hovering around 1 fps. The lightweight diagnostic frames take virtually no time to generate, and we have confirmed with logging that the AI generated frames in the production version are not generating any slower, and should still be able to meet the 18 fps demand. But we see that the time to send frames over websocket is on the order of 1s per frame, and is very unstable. See below a snippet from our logs showing fast image generation times, but slow times for sending images over websocket ```...
No description

Runpod down?

Getting error 400 for all our routes

On-Demand vs. Spot Pod

Hi ! I was read FAQ about its but I have one more question- is Spot Pod billing is also based on actual usage like on-demand? With power off like described of course....

Unable to access Custom Container & Community Templates

Hi team, I have credits on my RunPod account and have already deployed a pod, but I cannot find the “Custom Container” option nor any community templates in my UI. I’ve tried different browsers, cleared cache, and checked all menus (Templates, Explore, Pods), but the options simply don’t appear. Could you please check my account settings and unlock these features? Thank you!...

Ai Malware detection

Hi am a student majoring in cs, i would like to know your policies regarding malware detection ai/ml training on the pods, am i allowed to upload binaries and dll if not are malware features and hashes fine?

Facing a Serverless run error that is not encountered in ComfyUI web UI.

Hi all, my Serverless Endpoint has the same custom nodes and models as my persistent pod. However, it encountered the following error when I tried to run my custom workflow on the Serverless Endpoint. Please see logs attached😢 and help me to fix.
Solution:
I've set COMFY_POLLING_MAX_RETRIES to 1000 and it works

Bad requests

I have just the normal 3.6.0 sdxl docker running on my serverless but i just cannot get it to request correctly with a workflow. According to github which is literally the only place with any documentation it says a workflow can be passed but nothing I've tried works its just endless 400 bad request... ``` { "input": { "prompt": prompt,...

Run a function once when a worker starts

I have some code that I only want to run once when my serverless worker starts (i.e not everytime a request is made). What is the best way of doing this? Just executing the code outside of the handler? ``` import runpod import time...

Comfyui serverless via fastapi python to generate an image

Does anyone have a good python demo code to generate an image against a serverless comfyui . I am running a hackathon for a ai Microsoft project and desperately looking for fire base that works. Anyone willing to share some code base?...

Serverless VLLM concurrency issue

Hello everyone, i deployed a serverless vllm (gemma 12b model) through runpod ui. withj 2 workers of A100 80GB vram. if i send two requests at the same time, they both become IN PROGRESS but i recieve the ouput stream of one first, the second always waits for the first to finish then i start recieveing the tokens stream. why is it behaving live this?...

Bad requests

Hello I just started working on this I have an endpoint that's on just the default docker image that comes with the sdxl model. The issue is any request just ends up saying "[error]Failed to return job results. | 400, message='Bad Request'" Like I even copied the example request off the GitHub but it just won't do anything. I'm assuming I'm just doing the request wrong but there's no documentation stating otherwise....

Workers are Stuck on Booting Fetching Image - Anyone Else?

Experiencing workers stuck on booting and tries to fetch image indefinitely. I'm experiencing this on EUR-IS-2, if you experience it please share details + region.

Serverless vLLM changing engine arguments

Hi, I got vLLM Serverless worker up and running, but want to change one engine argument (which is not overridable through environment variables), specifically --limit-mm-per-prompt , how could I do that with your custom image runpod/worker-v1-vllm:v2.3.0stable-cuda12.1.0 that endpoints use? Thanks

Jobs stuck as no workers are being initialized

I have an endpoint that is not spinning workers anymore even when I try resetting the number of max/active workers.
No description

ComfyUI Job Failed

Based on the logs, it seems like it tries to connect before it's actually launched. Let me know if I should do anything on my end to fix. It worked well before....

why the hell are my delay times so high and im bearing all the costs??

yesterday everything was working fine, delay times were a couple seconds. but now the delay times are getting ridiculous and IM being charged for the delay on top of the execution??
Next