New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[OnlineRequestProcessor Enhancement] Better way to do output token estimation #206

Open

Tracked by

CharlieJCJ opened this issue Dec 4, 2024 · 0 comments

Contributor

CharlieJCJ commented Dec 4, 2024 •

edited by RyanMarten

Loading

For our client side rate limit control, we need to estimate the number of tokens that will be used by each request we send to an LLM completions API.

Right now we are naively estimating that the output of the call will use (max_output_tokens // 4).

Instead we should implement a moving average based on the responses we have gotten so far.

This will help with leaving extra throughput on the table as discovered when looking into #223.

CharlieJCJ mentioned this issue

OnlineRequestProcessor V2 Megathread #204

Open

4 tasks

RyanMarten mentioned this issue

Better token consumption estimation #235

Closed

CharlieJCJ mentioned this issue

Gemini Flash runs very slow (1 it/s) #223

Closed

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment