Extract_features form trained image net model.

User 2074 | 7/2/2015, 1:46:28 PM

Hi; I read the Dato blog "Deep Learning: Doubly Easy and Doubly Powerful with GraphLab Create." My note book got stuck (it looks like this "In [*]" for about 6 hours already) with this line of code: catsdogssf['features'] = pretrainedmodel.extractfeatures(catsdogssf) What should I do in order to fix this problem?

Many Thanks;



User 1178 | 7/6/2015, 4:08:23 PM

Hi GadH,

Can you tell us your computer configuration, do you happen to have GPU? As Piotr pointed out in his blog, most of the code in the blog will take quite a long time in a computer without GPU.

Can you also check if the computer is actively processing something while you are waiting? If not, then maybe there is some problem happened somewhere. If you have your python window still open, you may find where GraphLab Create server log is by finding some printout like:

[INFO] Start server at: ipc:///tmp/graphlab_server-3526 - Server binary: /Users/pingwang/dato-dev/debug/src/unity/python/graphlab/unity_server - Server log: /tmp/graphlab_server_1435876264.log Take a look at the server log and see if there is any error showing up.

If you still encounter problem, please send us the server log and we can take a look.