NO max pool layer after first convolution do affect the performance of whole model

[Jayan-K-Duggal]

I just came across this new thing that you don't have a max pool layer as proposed in baseline squeezenext after first convolution block.
Leaving this pooling layer out reduces the accuracy but better memory and model size.
Just for FYI!