Tutorials and example for CPU only mode of Caffe?

[Stian hotboi]

Dear someone,
I think I was able to install Caffe in a CPU only mode, and would like to test some of the tutorials.  However, it seems that the imageNet stuff require GPU and produces: ./build/tools/caffe train --solver=models/bvlc_reference_caffenet/solver.prototxt
I1107 18:45:55.073873  8207 caffe.cpp:184] Using GPUs 0
F1107 18:45:55.074015  8207 common.cpp:66] Cannot use GPU in CPU-only Caffe: check mode.
*** Check failure stack trace: ***
    @     0x7fc15405edaa  (unknown)
    @     0x7fc15405ece4  (unknown)
    @     0x7fc15405e6e6  (unknown)
    @     0x7fc154061687  (unknown)
    @     0x7fc1543ebbb0  caffe::Caffe::SetDevice()
    @           0x407961  train()
    @           0x4054a1  main
    @     0x7fc153570ec5  (unknown)
    @           0x405b8d  (unknown)
    @              (nil)  (unknown)
Aborted (core dumped)

Do anybody know how to modify the tutorial to only work on CPU? or does somebody know if any of the tutorials uses CPU only?

[Ninja At Work]

Hi

here you go

1-you may read the tuts link from main caffe site one more time specially 3 first

2-the problem you have with cpu will done if you change last line of the solver.prototxt ,solver type from GPU to CPU

3-i'll be glad if someone say's how long it take to do imagenet on cpu mode (simple core i7) but seems for now that isnt good choice for you,try by mnist and your own simple dataset

4-simple classification is to prepare the model architect for both train and val ,solver and a simple mean file,next you go by preparing the lmdb( level db or image folder) and related .txt containing categories then fire the train from caffe/tools. 

keep up the good work