سلام
من میخوام طبق یکی از مثال های متلب ویژگی های(feature) یکی از لایه های عمیق شبکه عصبی alexnet را با استفاده از روش فعال سازی استخراج کنیم
من توی اجرای این دستور:
featureLayer = 'fc7';
trainingFeatures = activations(net, trainingSet, featureLayer, ...
'MiniBatchSize', 32, 'OutputAs', 'columns');
با error زیر مواجه شدم :
Error using gpuArray
An unexpected error occurred during CUDA execution. The CUDA error was:
unspecified launch failure
در ضمن توی مٍثال این توضیحات رو هم گفته :
Note that the activations function automatically uses a GPU for processing if one is available, otherwise, a CPU is used. Because of the number of layers in AlexNet, using a GPU is highly recommended. Using a the CPU to run the network will greatly increase the time it takes to extract features.
In the code above, the 'MiniBatchSize' is set 32 to ensure that the CNN and image data fit into GPU memory. You may need to lower the 'MiniBatchSize' if your GPU runs out of memory. Also, the activations output is arranged as columns. This helps speed-up the multiclass linear SVM training that follows.
متلب من 2017 هست و cuda8 رو هم نصب کردم
ممکنه منو راهنمایی کنین