LdsConv: Learned Depthwise Separable Convolutions by Group Pruning
pytorch >= 1.1
Prepare your dataset Then just python train.py or python imagenet_train_densenet.py
| Model | Error(top-1) | GFLOPs | Params(M) |
|---|---|---|---|
| Lds-ResNet50 (k=2) | 22.9 | 2.71 | 16.8 |
| Lds-ResNet-extreme | 23.4 | 2.28 | 14.3 |