1 x 1 conv vs linear

[Ming-er]

What's the different between 1 x 1 and linear? why should we do the replacement?

[Vincent-luo]

@Ming-er I think the only difference between 1x1 conv and linear is that conv layer ignores the bias term in linear layer, and except that the calculation is the same. I'm curious about if the bias term will effect the final performance. Did you do some experiments to compare these two approaches?