【paddle】自带模型参数量和计算量统计

准备使用paddleslim的一些工具，顺便统计一下以便做模型优化时用到。
测试代码

import paddle
import paddle.vision.models as models

sides = [64, 128, 224, 256, 384, 512]
clses = [10, 100, 1000]
for side in sides:
    for cls in clses:
        net = models.vgg19(pretrained=False, num_classes=cls)
        # net = models.mobilenet_v1(pretrained=False, scale=1.0, num_classes=1000)
		# net = models.resnet34(pretrained=False)
        FLOPs = paddle.flops(net, input_size=[1, 3, side, side], print_detail=True)
1
2
3
4
5
6
7
8
9
10
11

Resnet18

Resnet34

Resnet50

mobilenet_v1

vgg13

vgg16

vgg19