gpuArray

由于我要用matcovnet 来训练CNN，但遇到了

Error using vl_nnconv

DATA and FILTERS do not have compatible formats.

经过查询，在GPU 上训练时，要使用 gpuArray 来进行格式转换。

函数介绍：

----------------------------------------------------------------------------------------

gpuArray

Create array on GPU

Syntax

G = gpuArray(X)

Description

G = gpuArray(X) copies the numeric array X to the GPU, and returns a gpuArray object. You can operate on this array by passing its gpuArray to the feval method of a CUDA kernel object, or by using one of the methods defined for gpuArray objects in Establish Arrays on a GPU.

The MATLAB array X must be numeric (for example: single, double, int8, etc.) or logical, and the GPU device must have sufficient free memory to store the data.

If the input argument is already a gpuArray, the output is the same as the input.

Use gather to retrieve the array from the GPU to the MATLAB workspace.

Examples

Transfer a 10-by-10 matrix of random single-precision values to the GPU, then use the GPU to square each element.

X = rand(10,'single');
G = gpuArray(X);
classUnderlying(G)

single

G2 = G .* G;         % Performed on GPU
whos G2              % Result on GPU

  Name       Size      Bytes  Class

  G2        10x10        108  gpuArray

Copy the array back to the MATLAB workspace.

G1 = gather(G2);
whos G1

  Name       Size      Bytes  Class

  G1        10x10        400  single