Could I use Convolutional Neural Network in Neural Network toolbox with GPU card of capabilty less than 3.0

1 visualización (últimos 30 días)
Image Category Classification Using Deep Learning example did not work with my Quadro 4000 card. It required GPU card with CUDA 3.0 or higher. Is it possible to adjust the code to use GPU card with lower capability?

Respuesta aceptada

Ben Tordoff
Ben Tordoff el 12 de Mayo de 2016
Hi Serghei, I'm afraid the answer is no. Neural Network Toolbox uses NVIDIA's cuDNN library for running Convolutional Neural Networks on the GPU and this library has always required a device with compute capability 3.0 or higher. As you have discovered, the Quadro 4000 is compute capability 2.0. Nearly all NVIDIA GPUs released since mid-2012 have compute capability 3.0 or higher.
  2 comentarios
RoiD
RoiD el 21 de Mayo de 2016
Hi, I'm also facing the same problem, I want to use trained-convNets as a feature extractor. Is there any possibilty to use Matlab toolbox with an older GPUs? or even running it on CPU instead?
(There is no need for extensive computational power unless you train a network from scratch)
-R
Chan  Yao Jun
Chan Yao Jun el 30 de Jul. de 2016
Hi, im facing the similar problem. The GPU in my laptop is GT540M and it has a compute capability of 2.1 only. Is there any other possible ways to run Convolutional Neural Networks in my laptop?
Thanks.

Iniciar sesión para comentar.

Más respuestas (0)

Categorías

Más información sobre Deep Learning Toolbox en Help Center y File Exchange.

Etiquetas

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by