Private Inference on Switch Base
3.04Runtime (min)SecMoE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SecMoENetwork Setting=WAN, Number of Experts=322026.01 | 3.04 | 2.09 | |
| BumbleBeeNetwork Setting=WAN, Number of Experts=322026.01 | 4.24 | 5.99 | |
| SecMoENetwork Setting=LAN, Number of Experts=322026.01 | 7.37 | — | |
| BumbleBeeNetwork Setting=LAN, Number of Experts=322026.01 | 7.88 | — | |
| SecMoENetwork Setting=LAN, Number of Experts=642026.01 | 7.88 | — | |
| IronNetwork Setting=LAN, Number of Experts=322026.01 | 8.91 | — | |
| SecMoENetwork Setting=WAN, Number of Experts=642026.01 | 11.72 | 2.18 | |
| SecMoENetwork Setting=LAN, Number of Experts=1282026.01 | 12.1 | — | |
| SecMoENetwork Setting=WAN, Number of Experts=1282026.01 | 14.73 | 2.91 | |
| BumbleBeeNetwork Setting=LAN, Number of Experts=642026.01 | 15.8 | — | |
| BumbleBeeNetwork Setting=WAN, Number of Experts=642026.01 | 17.05 | 9.98 | |
| IronNetwork Setting=WAN, Number of Experts=322026.01 | 17.4 | 28.51 | |
| IronNetwork Setting=LAN, Number of Experts=642026.01 | 19.2 | — | |
| BumbleBeeNetwork Setting=LAN, Number of Experts=1282026.01 | 32.3 | — | |
| BumbleBeeNetwork Setting=WAN, Number of Experts=1282026.01 | 34.89 | 19.13 | |
| IronNetwork Setting=LAN, Number of Experts=1282026.01 | 35.5 | — | |
| IronNetwork Setting=WAN, Number of Experts=642026.01 | 72.12 | 48.8 | |
| IronNetwork Setting=WAN, Number of Experts=1282026.01 | 143.78 | 86.68 | |
| BumbleBeeNetwork Setting=WAN, Number of Experts=162026.01 | — | 3.96 | |
| IronNetwork Setting=WAN, Number of Experts=162026.01 | — | 19.81 | |
| SecMoENetwork Setting=WAN, Number of Experts=162026.01 | — | 2.15 |