[GH-ISSUE #22] Relationship between parameters and qps ? #21

Open
opened 2026-10-26 21:10:20 +00:00 by bsakaguchi · 0 comments
Owner

Originally created by @Qia98 on GitHub (Dec 26, 2024).
Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-VL2/issues/22

In theDeepSeek-VL2-Tiny model, total activated parameters are 1.0B, but itis built on DeepSeekMoE-3B.

When inference, the inference parameter is 1b or 3b ?

If 1b, does it theoretically mean that its qps(Queries Per Second) is higher than other model like qwen2b?

Originally created by @Qia98 on GitHub (Dec 26, 2024). Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-VL2/issues/22 In theDeepSeek-VL2-Tiny model, total activated parameters are 1.0B, but itis built on DeepSeekMoE-3B. When inference, the inference parameter is 1b or 3b ? If 1b, does it theoretically mean that its qps(Queries Per Second) is higher than other model like qwen2b?
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-VL2#21
No description provided.