you model is wrong。

[PeiChaoXu]

the first inverted residual block is not using deepthwise separable convolution(groups=16)

[Dipendra Jha]

@afterimagex do you mean that line should be :
Conv2d(16, 16, kernel_size=(3,3), stride=(1,1), padding=(1,1), groups=1, bias=False)

[Octocat]

@afterimagex I think you are right, the code of InvertedResidual in this implementation is wrong. @d-li14