[GH-ISSUE #70] What's the Prompt and Response length in the Paper? #65

Open
opened 2026-10-26 21:07:23 +00:00 by bsakaguchi · 0 comments
Owner

Originally created by @JadeRay on GitHub (Jul 9, 2024).
Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-V2/issues/70

image

The paper has mentioned 50K tps for generation phase and 100K throughput for prefill phase, my question is what's the tested prompt length and repsonse length in avg and P99?

Originally created by @JadeRay on GitHub (Jul 9, 2024). Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-V2/issues/70 ![image](https://github.com/deepseek-ai/DeepSeek-V2/assets/7777102/67e3f117-4a83-4b59-85ac-364084535689) The paper has mentioned 50K tps for generation phase and 100K throughput for prefill phase, my question is what's the tested prompt length and repsonse length in avg and P99?
Sign in to join this conversation.
No labels
pull-request
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
deepseek-ai/DeepSeek-V2#65
No description provided.