These are chat archives for beniz/deepdetect

9th
Jan 2018
Yili Zhao
@panovr
Jan 09 2018 01:47
And if I my model directory name is yub_resnet50, which name should my model be?
The original model name is ResNet-50-model.caffemodel, but there was an error service imageserv mllib internal error: no model in /home/ylzhao/models/yub_resnet50 for initializing the net.
Emmanuel Benazera
@beniz
Jan 09 2018 06:26
This error can only occur when a model should be available and is not, i.e. after training or when trying to make prediction.
Yili Zhao
@panovr
Jan 09 2018 06:33
I test the same python script on two machines, one with cuda 8.0, one with cuda 9.1. The one with cuda 8.0 success, but the one with cuda 9.1 failed.
Should I submit an issue to deepdetect github?
Emmanuel Benazera
@beniz
Jan 09 2018 06:36
You may but out of the blue I'd say it's unlikely it's a bug, rather an issue in your configuration. In all cases, you need to prepare a gist with all calls to API and exact steps you are taking. Cuda 9.1 works fine for us.
Yili Zhao
@panovr
Jan 09 2018 06:37
OK. By the way, do you have any good methods to run dede as a daemon service in Ubuntu?
Emmanuel Benazera
@beniz
Jan 09 2018 06:44
you may want to use systemd, https://deepdetect.com/ami/systemd/
Yili Zhao
@panovr
Jan 09 2018 06:50
How about this one dede_start.sh? Does it like sudo ./dede -host 0.0.0.0 -port 5001?