Predict by CPU

[xurong1981]

I want to use CPU-only to make a edge detection by rcf, and I have revised Makefile.config to commet "USE_CUDNN := 1" and unconment "CPU_ONLY := 1", and re-build and re-install caffe. But when I run predict, one error occured like the followings:

---

F1115 15:03:29.547533 6127 cudnn_relu_layer.cpp:13] Check failed: status == CUDNN_STATUS_SUCCESS (1 vs. 0) CUDNN_STATUS_NOT_INITIALIZED

Is there anyone know how to solve such a problem ?
Besides, the model I fine-tuned is trained by GPU.

[yun-liu]

The default engine in prototxt is CUDNN. You can solve this problem by removing the engine setting: (engine: CUDNN).

[xurong1981]

Thank you for your reply. Actually, I changed "engine: CUDNN" by "engine: CAFFE", but it could not work. Is it available to remove ?

[yun-liu]

Yes, you can remove it. The default setting in caffe will decide whether to use CPU or GPU automatically.

[xurong1981]

I see, I will make a try. Thank you.

[yun-liu]

You are welcome!

[edWin-m]

Make sure you have installed cudnn correctly. i was having the same error before i fixed my cuDNN. Check the link below for the instructions.
http://docs.nvidia.com/deeplearning/sdk/cudnn-install/index.html#installlinux-deb