Deep Learning using GPU

Hi there,

I want to run my deep learning model (m script) for aesthetic designs (thousands of images of trochloids) using Quadro RTX 4000 GPU; i9 with 10 cores.

Any suggestion so I can run it efficiently? do I need to modify my m script?

Thanks,

S

Answer 1

[Jayanti]

Hi Sri,

If you want to run a deep learning model using a GPU in MATLAB, the first step is to verify whether GPU is supported. You can quickly check this by visiting the following link:

https://www.mathworks.com/help/parallel-computing/gpu-computing-requirements.html

If GPU is supported and you have the Parallel Computing Toolbox, MATLAB should automatically utilize the GPU for computations.

However, if GPU isn't supported, or if you are working with custom training loops, you will need to take an extra step. In this case, convert the data into “gpuarray” and ensure that you are using functions that support “gpuarray”.

Please refer to the below link to refer to the list of GPU supported functions:

https://www.mathworks.com/help/deeplearning/referencelist.html?type=function&capability=gpuarrays

For more detailed guidance on running deep learning models on a GPU, you might find the below documentation helpful:

https://www.mathworks.com/help/deeplearning/ug/train-network-using-custom-training-loop.html#:~:text=128%3B%0AlearnRate%20%3D%200.01%3B-,Train%20Model,-Create%20a%20minibatchqueue

Hope this will be helpful!