Pruning Strategy Search on ImageNet
25GPU HoursEagleEye
Evaluation results
| Method | Links | |
|---|---|---|
| EagleEyeEvaluation Method=adaptive-BN, Candidate Selection=<1000x100 inference iterations2020.07 | 25 | |
| NetAdaptEvaluation Method=finetuning, Candidate Selection=10^4 training iterations2020.07 | 864 | |
| ThiNetEvaluation Method=finetuning, Candidate Selection=1000x10 finetune epochs2020.07 | 8,000 | |
| Filter PruningEvaluation Method=vanilla, Candidate Selection=1000x25 finetune epochs2020.07 | 20,000 |