2

Using the built-in graphics cards without a NVIDIA graphics card, Can I use the CUDA and Caffe library ?

My O/S is ubuntu 15 CPU is (Intel)i5-4670(3.40GHz 4 core) RAM is 12.0GB

I want to start to study deep learning.

Ciro Santilli OurBigBook.com
  • 347,512
  • 102
  • 1,199
  • 985
Junsuk Lee
  • 57
  • 7
  • 1
    You may be interested in [this question/answer](http://stackoverflow.com/questions/31688759/mac-caffe-cuda-driver-issue) – Robert Crovella Apr 05 '16 at 02:19
  • GeForce GTX 750 Ti with 2GB memory. About US$130, easily available in most of the world. This will run a lot of the examples and tutorials. The big VGG nets need more memory. You can often reduce batch size to get around the lack of GPU memory. I have found it great for learning and testing, much faster than CPU_ONLY. – cgarner Apr 06 '16 at 13:14
  • Possible duplicate of [GPU Emulator for CUDA programming without the hardware](http://stackoverflow.com/questions/3087361/gpu-emulator-for-cuda-programming-without-the-hardware) – Ciro Santilli OurBigBook.com Feb 16 '17 at 17:57

1 Answers1

3

CUDA is for NVidia graphics cards. Therefore you cannot make use of the CUDA built of Caffe in your PC.

Still, you could make use of the CPU_ONLY flag during built and get Caffe working on your system. But you can expect the training to be too slow compared to what you should get when using a graphics card.

There are various fork projects of Caffe which are made to make it work on other graphics card and are available on Git. But most of these projects are seriously outdated when compared with the master branch of original Caffe.

An OpenCL implementation of Caffe can be found here.

Anoop K. Prabhu
  • 5,417
  • 2
  • 26
  • 43