Pytorch统计网络参数
def get_parameter_number(net):
total_num = sum(p.numel() for p in net.parameters())
trainable_num = sum(p.numel() for p in net.parameters() if p.requires_grad)
return {'Total': total_num, 'Trainable': trainable_num}
print(model.state_dict())
FLOPs, MACs, MAdds 关系
见文章:CNN模型复杂度(FLOPs、MAC)、参数量与运行速度
计算工具:
地址备注https://github.com/Lyken17/pytorch-OpCounterPytorchhttps://github.com/sovrasov/flops-counter.pytorchPytorchhttps://stackoverflow.com/questions/45085938/tensorflow-is-there-a-way-to-measure-flops-for-a-modelTensorFlow: 自带tf.RunMetadata()
另:在PyTorch中,可以使用 torchstat这个库来查看网络模型的一些信息,包括总的参数量params、MAdd、显卡内存占用量和FLOPs等。
!pip install torchstat
from torchstat import stat
from torchvision.models import resnet50, resnet101, resnet152, resnext101_32x8d
model = resnet50()
stat(model, (3, 224, 224))
total = sum([param.nelement() for param in model.parameters()])
print("Number of parameters: %.2fM" % (total/1e6))
也可以使用 torchsummary
!pip install torchsummary
from torchsummary import summary
summary(model, input_size=(ch, h, w), batch_size=-1)
Original: https://blog.csdn.net/user_lib/article/details/123452572
Author: 李代数
Title: Pytorch统计网络参数计算工具、模型 FLOPs, MACs, MAdds 关系
原创文章受到原创版权保护。转载请注明出处:https://www.johngo689.com/710574/
转载文章受原作者版权保护。转载请注明原作者出处!