Nateliu712

Nice, work. I wonder to know if this can work in Seq2Seq model, like attention based model.

@mialo-tech I meet the same question. Do you fix the problem?

I met the same problem. And I compare this code with the https://github.com/jaxony/ShuffleNet and the paper. I find the 3*3 convolution in the bottleneck unit should be depthwise convolution which...