Tian, Feng

I reproduced your issue at my side. I am digging into it. it looks like a memory corruption issue caused by mkldnn. will get back to you if I have...

sorry for late response. I ran your test code on SKX, the error I met is a little different with yours, which crash on posix_malloc() with unresonable size. After digging...

you can refer to how examples/faster-rcnn/lib/datasets/pascal_voc.py _load_pascal_annotation() handles VOC2007 annotation xml file.

are you sure that? intelcaffe allocate thread number according to phyical core number if no OPNEMP env variable setting. pls be aware of the difference between physical core and logical...

MKL2017 is deprecated and should be not used. as for net_->Reshape(), it will be invoked at net_->Forward(). your code makes this func invoked twice and bring state machine wrong. (The...

@yflv-yanxia could you let me know the detail steps of reproducing the accuracy issue?

alexnet contains fullyconnect layer, it doesn't allow variable batch size.

fully connection layer weight number usually is oc x ic x ih x iw if the axis is 1. if the axis is 0, the weight number would be oc...

thanks @lddsdu . we usually do same thing by having a python layer. it's mainly because user may not know the time step before data preprocessing. so it's hard to...

you can have a try my repo https://github.com/ftian1/intelcaffe_weight_to_bvlc