code and results for <<Pruning Filters For Efficient ConvNets>>
Third party: https://github.com/Eric-mingjie/rethinking-network-pruning/tree/master/imagenet/l1-norm-pruning
Metric: Weight value
Network:
vgg16 rc56(resnet 56)
GPU:
batch: 1 32 64 128
pruning ratio: 0.2 0.4 0.6 0.8
environment: Tensorrt
precision: FP16 INT8
DLA:(max 32)
batch: 1 8 16 32
pruning ratio: 0.2 0.4 0.6 0.8
environment: Tensorrt
precision: FP16 INT8
CPU:
batch: 1 32 64 128
pruning ratio: 0.2 0.4 0.6 0.8
environment: onnxruntime
precision: FP32