查模型的参数Param和浮点计算量FLOPs

    from thop import profile
    
    input = torch.randn(1, 1, 64, 64).cuda()
    model = GroupMixFormer().to('cuda')
    
    model.eval()
    output = model(input)
    flops, params = profile(model, inputs=(input,))
    print("input_shape:", input.size())
    print("output_shape", output.size())
    print('   Number of parameters: %.2fM' % (params / 1e6))
    print('   Number of FLOPs: %.2fG' % (flops * 2 / 1e9))