tryolabs / luminoth

Deep Learning toolkit for Computer Vision.
https://tryolabs.com
BSD 3-Clause "New" or "Revised" License
2.4k stars 400 forks source link

First trial and OOM #258

Open jmFschneider opened 5 years ago

jmFschneider commented 5 years ago

Hello,

I am just starting with with Luminoth and try "lumi predict image.jpg"

I am on a PC under unbutu with 8 giga octet of ram, a GPU with 2 giga and installed tensorflow-gpu. The detection did not complet and I have the following message "ResourceExhaustedError (see above for traceback): OOM when allocating tensor with shape[2000,14,14,1024] "

Withing Keras, I was able to modify the code to avoid this error. Is it to do some think here ?

Thnaks you for your help Jean-Marie

meyerjo commented 5 years ago

You could try to reduce the post_nms_top_n in the config file which is by default 2000.