ShuffleNet icon indicating copy to clipboard operation
ShuffleNet copied to clipboard

Message type "caffe.LayerParameter" has no field named "shuffle_channel_param"

Open XingyuXie opened this issue 8 years ago • 15 comments

I have added the shuffle layer into caffe, and run 'make all' successfully. And I have trained the model with shuffle layer successfully. But when I run the test code in C++, there is an error 'Error parsing text-format caffe.NetParameter: 109:25: Message type "caffe.LayerParameter" has no field named "shuffle_channel_param". I have already added it in caffe.proto, why this problem still exists? Thanks a lot.

XingyuXie avatar Aug 04 '17 07:08 XingyuXie

when i use draw_net.py to draw this net ,also have this error:Message type "caffe.LayerParameter" has no field named "shuffle_channel_param" and make successfully, hope your reply. thanks a lot.

yanyanxixi avatar Aug 11 '17 06:08 yanyanxixi

@yanyanxixi Have you make pycaffe successfully?

farmingyard avatar Aug 11 '17 06:08 farmingyard

@farmingyard thanks a million

yanyanxixi avatar Aug 11 '17 07:08 yanyanxixi

I have the same propblem. Have u solved it ?

CyrilYang avatar Aug 11 '17 09:08 CyrilYang

想问下,shuffle Channel 具体的操作是怎样的,看了论文也不太清楚具体的操作,是直接任意打乱再均分到下一层不同group,还是有其它的做法,还有就是反传的时候会对shuflle Channel 有什么调整吗,谢谢

CLAPoo avatar Aug 23 '17 05:08 CLAPoo

@CyrilYang The model still cannot run on the docker. But there is no problem on my own ubuntu. So I ran it on my local ubuntu instead.

XingyuXie avatar Aug 23 '17 09:08 XingyuXie

@XingyuXie Thanks a lot!

CyrilYang avatar Aug 23 '17 10:08 CyrilYang

@XingyuXie 你好,请问在训练ShuffleNet时是不是需要 下面内容添加到caffe.proto中然后重新编译caffe? optional ShuffleChannelParameter shuffle_channel_param = 164; }

message ShuffleChannelParameter { optional uint32 group = 1[default = 1]; // The number of group }

lrjun6188 avatar Sep 05 '17 07:09 lrjun6188

@LiuRJun 是的

XingyuXie avatar Sep 08 '17 01:09 XingyuXie

@farmingyard @yanyanxixi how did you deal with error: :Message type "caffe.LayerParameter" has no field named "shuffle_channel_param"? I make pycaffe successfully, but when I test the pre_trained model with python I met this error. I need your help.

houxiaoxia1 avatar Nov 14 '17 06:11 houxiaoxia1

my problem is about pycaffe. not make, can you use draw_net.py to draw this net successfully? @houxiaoxia1

yanyanxixi avatar Nov 14 '17 07:11 yanyanxixi

I can draw this net successfully. but I can't use the deploy.prototxt with to test picture. my code: import caffe import glob import os import cv2

WEIGHTS_FILE = '/home//caffe/shufflenet_step2_iter_5000.caffemodel' DEPLOY_FILE = '/home//caffe/models/shufflenet/shufflenet_1x_g3_deploy.prototxt' I met the error when I test a picture. @yanyanxixi @farmingyard

houxiaoxia1 avatar Nov 14 '17 07:11 houxiaoxia1

Hi, did you guys solve this problem? I can train this network successfully. However, when i do testing, i have the same problem:

Message type "caffe.LayerParameter" has no field named "shuffle_channel_param".

Thanks

DanChen001 avatar Sep 07 '18 16:09 DanChen001

@farmingyard I can train this network successfully. However, when i do testing, i have the this problem: Message type "caffe.LayerParameter" has no field named "shuffle_channel_param".

do you know how to solve this problem? thanks

DanChen001 avatar Sep 07 '18 17:09 DanChen001

@DanChen001 did you solve this problem?

miaomi1994 avatar Nov 14 '18 09:11 miaomi1994