mitmul / deeppose

DeepPose implementation in Chainer
http://static.googleusercontent.com/media/research.google.com/ja//pubs/archive/42237.pdf
GNU General Public License v2.0
408 stars 129 forks source link

Memory problem #32

Open gingerhead22 opened 7 years ago

gingerhead22 commented 7 years ago

Anyone has an idea how much memory is in need to run this program? I have 16GB and it seems like that is way not enough. Thanks

quanweikikai commented 7 years ago

you mean GPU memory?

you can modify batch size to control memory consuming. Usually, this one will not take that much memory in my case.

gingerhead22 commented 7 years ago

Nope, I already have the GTX 1070. I mean the regular memory, cos' it is reported no memory after running 20mins. Did the code work fine on your computer? Thank you!

quanweikikai commented 7 years ago

The code did not run on my PC but I customized it and add the TEST phase to the network so it work well. If you need I can give you as a reference. It seems that your case is more like other problem but not the network or caffe problem. I guess maybe some code did not release the image in memory(maybe opencv?) or something like that. By the way, I only used the caffe branch.

gingerhead22 commented 7 years ago

Thank you for you comment. I ran it with chainer on windows 10, so it may caused by system problem. I would be so nice of you if you can send me yours as a reference. Thank you very much then.