Find answers from the community

Updated 3 months ago

hey guys building a small app with the

hey guys ! building a small app with the latest version, I notice it's not printing out the token usage in the console , it used to do that in an older version ....

5 comments

LLogan M

The prints are still there, just disabled by default for now.

You can re-enable logs with this:

Plain Text

import logging
import sys

logging.basicConfig(stream=sys.stdout, level=logging.INFO)
logging.getLogger().addHandler(logging.StreamHandler(stream=sys.stdout))

Very soon too (i.e. the next release), the token counting will be available as a callback, for more programmatic access to token counts

JJay

Ok awesome thanks

JJay

I noticed if I switch to gpt-turbo-3.5 , it is a LOT slower than davinci-text-003 , super weird. I'm just trying to save on cost, any idea if this is normal ?

JJay

it takes 8-15 secondes to run a query vs 2-3 seconds with davinci...

LLogan M

I think the server demand for gpt-3.5 is likely a lot higher... but also maybe OpenAI has shifted resources to those new openai models too: "gpt-3.5-turbo-0613" and "gpt-4-0613"

Add a reply