版权声明：xbx https://blog.csdn.net/weiweixiao3/article/details/82319416
Layer protocol
setup：run once for initialization
forward: make output given input
backward: make gradient of output
-w.r.t. bottom
-w.r.t. parameters(if needed)
reshape: set dimensions
Blob: 存储信息，
1. data, derivatives, and parameters
2. 分配内存
3. shuttle betwwen cpu and gpu
resume training
–snapshot=xxxxxx.solverstate
fine-tune
Lizi:
python examples/finetune_flickr_styl/assemble_data.py --workers=1 --imags=80000 --seed 831486
./script/download_model_binary.py model/bvlc_reference_caffenet`
./build/tools/caffe train -solver models/finetune_flickr_style/solver.prototxt \
-weights models/bvlc_reference_caffenet/bvlc_reference_caffenet.caffemodel -gup=0Inference
需要对prototxt修改下，data layer
