Spaces:
Running
Infinite Loading
Hello and thank you for the amazing project.
I have been trying to access this Space, but it seems that Gradio is stuck in an infinite loading loop. I have tried different browsers, but there is no difference.
Seems it is working now! Closing this...
This issue has been happening for multiple days now, so it should be looked into.
@analogspiderweb Let's reopen this, then.
Same here
same
doesnt load
Same for me. It has been a few days.
same here
The creator probably doesn’t have much use for countless comments “same here”. You would help him by providing screenshots or other details of the error you encounter. For me, that means sometimes a full white screen, with nothing else. Other times, it seems stuck in a “loading” state. Could it be the endpoint is overloaded? This container only seems to run a simple nginx proxy, so that should be fine with any load thrown at it. Anyways, I’ve used this Whisper JAX a week ago without issue and am truly amazed by the speed 😳.
Great work Sanchit!
Should be fixed - sorry about that all! Really difficult to maintain this space tbh - hard linking up a TPU back-end to a Gradio CPU front-end, so many small issues that break the demo. Getting better at monitoring it and catching, but these pings are useful to know when it goes down, so thank you all!
@sanchit-gandhi it was working a couple of hours ago, but now it seems stuck in loading again.
@sanchit-gandhi it was working a couple of hours ago, but now it seems stuck in loading again.
..and now it works again!
It is still down :(
Hmm, no logs for me this time 🤔 rebooted the TPUs and it seems back up again, feel free to ping me if it goes down! Seems like my monitoring script didn't catch that one
Hey, thank you for your answer,
However I encounter a problem where the loading time would go up to 4000+ seconds on the queue (and/or right after)and it would never stop processing, do you know any way to fix it please?
The audio file I'm trying to upload is about 8mins long.
By the way, do you know how to deal with bilingual documents ? To specify two languages in the same file?
Thank you!
That sounds like a problem with the demo hanging - feel free to post on here when the loading time goes over a reasonable amount (e.g. for an 8 mins audio it shouldn't be more than a few tens of seconds)
do you know how to deal with bilingual documents
That's a good question! As far as I'm aware this is a limitation of the Whisper model in that it makes it's prediction for a single language per 30s segment based on the language tag predicted
Might be some ideas here! https://github.com/openai/whisper/discussions/49