Huggingface cached models seems to not working
i`m added repo but no models inside container and no logs
Serverless Worker Crashed but Request Still Running
A serverless worker suddenly died. However, an inexplicable phenomenon occurred — the request is still being processed.
Endpoint ID: h10qsr3s6f5puk
Request ID: 1c30dc85-d5e4-472d-b5e8-034d40249e7c-e2
Worker ID: 01yuqqrddjl88x...
Nvidia-smi parsing error
Hi,
I’m encountering an issue when launching a video generation on my RunPod instance.
Sometimes, this error message appears repeatedly:
...
Workflow TO Api wizard not working properly
I've run 3 different docker images, and it keeps saying build failed but i don't know why it doesn't give me any error messages
Can anybody help?...
RuntimeError: CUDA driver initialization failed, you might not have a CUDA gpu.
I am using CUDA 12.9, and I have also set "Allowed CUDA Versions" to 12.9, but sometimes an error occurs. Could you please tell me how to solve this problem? logs show RuntimeError: CUDA driver initialization failed, you might not have a CUDA gpu.
Serverless loadbalancer scaling config (Request Count)
From the documentation:
"Request Count scaling strategy adjusts worker numbers according to total requests in the queue and in progress. It automatically adds workers as the number of requests increases, ensuring tasks are handled efficiently. Total Workers Formula: Math.ceil((requestsInQueue + requestsInProgress) / 4)
Use this when you have many requests and workers won't have a chance to idle (e.g., with vLLM). This allows your app to scale down when traffic drops. With queue delay, once a worker scales up, if it's always busy, which makes scaling down harder."
It's not clear at what interval Math.ceil, runs at. If my server can handle 10 requests per second, what's a good value for the Request Count? Also, since my FastAPI handels an internal Queue, requestsInQueue will always be 0.
...
Serverless FAILING to add Workers
I have a queue-based endpoint created & i have 4 requests in the pipeline.
It's been over 30-40 mins and Serverless has failed to recruit any new H100 worker for me.
I don't have any data-centers (regions) specified....
