the paper with details: https://arxiv.org/pdf/2411.02265
They use
- 16 experts, of which one is activated per token
- 1 shared expert that is always active
in summary that makes around 52B active parameters per token instead of the 405B of LLama3.1.
They use
- 16 experts, of which one is activated per token
- 1 shared expert that is always active
in summary that makes around 52B active parameters per token instead of the 405B of LLama3.1.
No comments yet.