Encrypted Language Model Inference on GLUE RTE
8.8Execution Time (s)Cerium
Evaluation Results
| Method | Links | |
|---|---|---|
| CeriumGPU=B200, Parallelization=8x2025.12 | 8.8 | |
| CeriumGPU=H100, Parallelization=8x2025.12 | 10.6 | |
| CeriumGPU=B200, Parallelization=4x2025.12 | 10.8 | |
| CeriumGPU=H100, Parallelization=4x2025.12 | 13.3 | |
| CeriumGPU=B200, Parallelization=2x2025.12 | 19.6 | |
| CeriumGPU=A100, Parallelization=8x2025.12 | 19.6 | |
| CeriumGPU=H100, Parallelization=2x2025.12 | 23.91 | |
| CeriumGPU=A100, Parallelization=4x2025.12 | 24.6 | |
| CeriumGPU=B200, Parallelization=1x2025.12 | 28.3 | |
| CeriumGPU=H100, Parallelization=1x2025.12 | 36.1 | |
| CeriumGPU=A100, Parallelization=2x2025.12 | 44.3 | |
| CeriumGPU=A100, Parallelization=1x2025.12 | 66 | |
| THORGPU=A100, Parallelization=1x2025.12 | 602.1 |