FLOPS

[sjfei]

大佬想问一下，FLOPS怎么知道的

[Lebhoryi]

谷歌一下，网上都能查到，一个是自己计算，一个是所用的神经网络框架自己计算得到

[noahzhy]

you can get the flops of you model as follows:

    from ptflops import get_model_complexity_info

    h, w = (192, 192)

    with torch.cuda.device(0):
        macs, params = get_model_complexity_info(
            model, (3, h, w),
            as_strings=True, print_per_layer_stat=True, verbose=True
        )
        flops = str(float(macs.split(' ')[0]) * 2) + " GFLOPs"
        print('{:<30}  {:<8}'.format('Computational complexity: ', macs))
        print('{:<30}  {:<8}'.format('FLOPs: ', flops))
        print('{:<30}  {:<8}'.format('Number of parameters: ', params))