caffe-builder icon indicating copy to clipboard operation
caffe-builder copied to clipboard

Hi, I want to know how to add an new layer type to caffe(python version)(install by anaconda)

Open guanhuankang opened this issue 4 years ago • 0 comments

I install caffe in my anaconda envs by running the following commands: conda create -n caffe python=3.7 conda install caffe-gpu [-c willyd]

Actually, I install caffe successfully. But I try to test a model download from someone's github. I meet new problems since the model I want to run exist some new layer type( layer:IRNN). And you can find the irnn.hpp and its relevant information in https://github.com/xw-hu/CF-Caffe. But I don't know how to add this new feature into the existing environment. Thank you for any help!

Reproduce

  1. install anaconda3
  2. install caffe with anaconda3 by running the above commands
  3. try to load the model(https://github.com/xw-hu/DSC/blob/master/examples/DSC/DSC_detection/deploy.prototxt). Yeap, I want to do some test on this model.
  4. you may found the following error msg:
W0822 23:59:18.000824 10196 _caffe.cpp:175] DEPRECATION WARNING - deprecated use of Python interface
W0822 23:59:18.000824 10196 _caffe.cpp:176] Use this instead (with the named "weights" parameter):
W0822 23:59:18.000824 10196 _caffe.cpp:178] Net('deploy.prototxt', 1, weights='snapshot/DSC_iter_12000.caffemodel')
[libprotobuf ERROR C:\ci\libprotobuf_1523040574637\work\src\google\protobuf\text_format.cc:288] Error parsing text-format
caffe.NetParameter: 524:15: Message type "caffe.LayerParameter" has no field named "irnn_param".
F0822 23:59:18.018841 10196 upgrade_proto.cpp:90] Check failed: ReadProtoFromTextFile(param_file, param) Failed to parse NetParameter file: deploy.prototxt
*** Check failure stack trace: ***

Thank you for any help! I want to know how to add IRNN to my envs.

guanhuankang avatar Aug 22 '20 16:08 guanhuankang