MATLAB Deep Learning Training with GPU using Neural Networks

Hi Guys, I make my question very simple, I hope getting answers.

I have a dataset of about 65GB having 1 lac plus images. I want to run deep lerning training using Neural Networks. Problem is to run such large dataset, my laptop took almost 30 plus days as processing of large dataset is difficult without fast GPUs. Suggest me best way to do training my model with high speed either its paid or free. I am looking much needed answers. Thnaks.

Answer 1

[David Willingham]

Hi Mohammad,

As you have pointed out with your application, some deep learning training tasks can take a long time. My suggestion would be to see if any of the virtual machines on AWS or Azure will help you get your results sooner.

The MATLAB Deep Learning Container can be used to install MATLAB quickly and easily on any of the virtual machines from any cloud provider.

I'd recommend:

1. Selecting a virtual machine with a GPU, such as Amazon's P3 Instances.
2. Pull & Run the MATLAB Deep Learning Container.
3. See how long it takes to execute ~10-100 epochs (rather than waiting multiple days).
4. See if this is quicker than your current desktop.

Regards,

David Willingham

Deep Learning Product Manager

[MOHAMMAD AQEEL]

Hi David, I have tried to follow same as you recommended. But the NGC container but it says that i don't have license for products MATLAB Coder and GPU Coder. I have make request to sales support for quote increase but didn't get yet. I am looking for response. Thank you.

[David Willingham]

Hi Mohammad, you shouldn't require a licence for all the products in the container. When you launch the VM and log into MATLAB only your licensed products will be available.

David

[MOHAMMAD AQEEL]

when i enter Mathwork Acount Username and Password, it says that you need license for GPU Coder and MATLAB Coder. I will attach screenshot next time very soon when i will try to launch the container. I have trial licnese for MATLAB and it says you need qoute increase for your trial. I will attach screenshot for clarity.