Error Using trainNetwork in Deep Learning.

Hi,

I am trying to train a pretrained CNN model, inceptionresnetv2 for Image Classification. I am getting the following error,

Error using trainNetwork (line 150)

Error using 'predict' in Layer nnet.inceptionresnetv2.layer.ScalingFactorLayer. The function threw an error and could not be executed.

Any solution??

Thank you!!

[Gabija Marsalkaite]

Hi Kaniska,

Could you provide more information about the problem e.g. hardware used (GPU and amount of RAM) and if that is the full error message?

[Kaniska Samanta]

Hi Gabija,

My PC configuration is Processor: Intel Core i5 (6th Gen); RAM: 8GB; GPU: Nvidia 930M; Storage: 240GB SSD.

And yes that was the full error message.

But now I am facing a new error sometimes as shown in the picture below, and I am getting this error after 4 or 5 iteration in the training progress.

[Maria Duarte Rosa]

Using a smaller network (e.g. squeeznet or mobilenetv2) and reducing the mini-batch size might help.

[Gabija Marsalkaite]

Hi Kaniska,

Most likely your GPU is either low on memory or out of memory. I don't know the exact settings you are using but reducing mini-batch size or using CPU instead of GPU may help (CPU would use 8GB of RAM instead of GPU using 2GB of its memory).

[Kaniska Samanta]

Hey Gabija Marsalkaite,

Thanks for assists. It's propably my gpu, unable to handle the network..

[Ansuman Mahapatra]

Reduced batch size to 16 and it works.