Private Inference on T5 138M
7.62Embed Inference Time (s)MERGE
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| MERGEOptimization Strategy=ER+MM2023.05 | 7.62 | 1.27 | 131.31 | 44.11 | 0 | 0 | 144.02 | 10.89 | |
| MERGEOptimization Strategy=only ER2023.05 | 8.24 | 1.58 | 211.57 | 65.19 | 596.74 | 166.5 | 874.36 | 1.79 | |
| THE-X2023.05 | 316.16 | 48.58 | 321.9 | 90.82 | 126.73 | 25.51 | 1,050.28 | 1.49 | |
| MERGEOptimization Strategy=only MM2023.05 | 322.38 | 51.35 | 221.57 | 69.22 | 0 | 0 | 693.3 | 2.26 | |
| CrypTen2023.05 | 323.46 | 53.36 | 328.09 | 96.08 | 693.73 | 175.57 | 1,569.41 | 1 | |
| MPCformerSoftmax Approximation=sm2quad2023.05 | 324.81 | 52.03 | 325.97 | 92.89 | 235.54 | 58.47 | 1,149.07 | 1.37 | |
| MPCformerSoftmax Approximation=sm2relu2023.05 | 327.51 | 55.36 | 328.61 | 96.8 | 284.65 | 75.17 | 1,207.63 | 1.3 |