yaozhewei / hap Goto Github PK

Hi, thanks for the code! Regarding the residual connections, I couldn't find the section of code that handles the structured pruning of these connections. I'm probably missing something, could you point me toward this? TIA.

请问论文中的这个公式

最后一项为什么是三次方啊？
按照最优化方法中的公式，感觉最后是个二次方啊....

Hi,
I'm trying to run the code and there is a missing python file. In line 17 and 18 of hessian_pruner.py there are two functions that are called from pyhessian but there is no such file!
Please let me know if I'm missing something.

Hi, thanks for your cool work.

An important related work maybe missed.

Collaborative Channel Pruning for Deep Networks. ICML2019

CCP also uses Taylor expansion to calculate the change of loss with respect to pruning. And approximated Hessian information is also used to measure the pruning sensitivity.