Spaces:
Running
on
L40S
Facial Expression Editor
Hi
Your facial expression editor has been broken for one week.It shows having large queues.Will you be fixing this?Thank you
Please fix the constant queue problems because the service is down. You will have a lot of people thanking you as the service is great. Thank you in advance
Yes please, it would help me and a lot of fans, thanx
Fixed by rebuilding, i think the redundant failing queue might be caused from cog server not able to handle too much too fast requests.
What do you think ?
OK, if that happens, maybe you can think about adjusting your error handling. It might be good to investigate what error message is coming from the cog-server! Is that a good idea?
here is the last significant change made, let’s see how tha app now handles queue / cog errors:
"Previously, the Gradio app would poll the Cog prediction status indefinitely, waiting for a "succeeded" or "failed" response. If the Cog server became unresponsive or a prediction hung (e.g. due to a bad input or internal model error), the polling loop would never exit.
This behavior caused the Gradio queue to back up and eventually stall completely under high load, since all workers could get stuck waiting for unresponsive predictions.
This change introduces a timeout mechanism (60 seconds) in the polling loop. If the prediction doesn't complete within that time, an error is raised and the request is gracefully terminated. This helps avoid queue deadlocks and improves reliability under concurrent usage."
thnx for notifying, let's see if this method helps to avoid queue issues.
May the force be with you
Grtz Erik