search

WeixiangXu / STTN

17 stars 0 forks source link

Why is the performance in this repo much better than paper? #1

Closed ThisisBillhe closed 10 months ago

ThisisBillhe commented 1 year ago

any tricks or reasons?

Cyber-Neuron commented 1 year ago

any tricks or reasons?

I think the avgpool, trining epochs, and preBN are some of the tricks.

WeixiangXu commented 10 months ago

sorry for late reply, the link may help. https://github.com/HolmesShuan/Training-Tricks-for-Binarized-Neural-Networks