lot of error 500s when using functions | OpenAI | Page 1

sour condor Jun 16, 2023, 4:24 PM

#

Hi all, shortly after I got my application using the new functions interface smoothly, after a few successful tests, the server suddenly started returning error 500 from almost every API call. It is possible to get support to help diagnose this?

vestal barn Jun 19, 2023, 4:44 PM

#

Yeah, I'm receiving 500s right now too.. Seemed to start when I started using nested objects, but it may be the same issue

rustic sail Jun 19, 2023, 5:14 PM

#

Same here but half of my normal calls coming back too 500 {'error': {'message': 'The server had an error while processing your request. Sorry about that!'

noble sleet Jun 19, 2023, 7:07 PM

#

Saw a 500 error yesterday even without function calls, something's up with the new model

quiet cape Jun 19, 2023, 7:17 PM

#

noble sleet Saw a 500 error yesterday even without function calls, something's up with the n...

Confirmed.

rancid smelt Jun 19, 2023, 8:40 PM

#

I got lots of error 500 in the past 3 days for using it normally without function calling...

rancid smelt Jun 19, 2023, 10:04 PM

#

just joined the party about an hour ago - only for gpt-4-0613, gpt-3.5-turbo-16k is fine

quiet cape Jun 19, 2023, 11:23 PM

#

I change to use gpt-4-0314, it's stable.

rancid smelt Jun 20, 2023, 3:34 AM

#

Ah yes, that solves the problem. I guess gpt-4-0613 isn't stable yet...

last jolt Jun 20, 2023, 6:13 AM

#

I see the same problem. gpt-3.5-turbo-16k-0613 works fine with functions.

chrome crater Jun 20, 2023, 1:54 PM

#

gpt-3.5-turbo-16k-0613 too giving 500s for me

severe hamlet Jun 21, 2023, 11:01 AM

#

quiet cape I change to use gpt-4-0314, it's stable.

Does it have functions?

#

Still getting 500s on 4-0613

jovial gate Jun 21, 2023, 11:07 AM

#

Since starting to use function calling I've also been getting "The server had an error while processing your request" limits. Has anybody already figured out how to sort it?

severe hamlet Jun 21, 2023, 2:52 PM

#

severe hamlet Still getting 500s on 4-0613

it seems to be fixed now

#

maybe just gpu prioritisation stuff idk

#

nvm this didnt age well

undone lava Jun 21, 2023, 6:20 PM

#

same getting 500s

#

tried a couple days ago and it worked tho

tame sentinel Jun 21, 2023, 11:07 PM

#

500s should be completely independent of your code. It means the server itself is having a problem. 500s have been constant with OpenAI APIs for months. Think of it like weather....

tranquil bison Jun 23, 2023, 1:12 AM

#

is everyone still seeing these 500s?

#

much much better today!

severe hamlet Jun 23, 2023, 6:14 AM

#

tranquil bison much much better today!

Yeah it’s gotten better recently

undone lava Jun 23, 2023, 8:23 AM

#

tame sentinel 500s should be completely independent of your code. It means the server itself ...

Then why does the same exact code work on 3.5 turbo?

#

Havent tested on 4 again since then, ill try again

#

But yea, same exact code, just changing the model gave a 500 error

#

Not always but over half the time

tame sentinel Jun 23, 2023, 2:21 PM

#

undone lava Havent tested on 4 again since then, ill try again

Even if API gateway is the same via the /chat/completion endpoint, I'm pretty sure the model parameter dispatches the call to different services. IF the GPT-4 service is unavailable, I would expectd it to generate a 500 error, which would bubble up to the API gateway and then back to the calling client. If the GPT-3 service is availble, then it would work without that error.

undone lava Jun 23, 2023, 2:30 PM

#

nvm it was early in the morning and i read "should be completely dependent of your code" shame