MATLAB: Do I get an error when I try to train a DAG network on multiple GPU’s

Deep Learning ToolboxParallel Computing Toolbox

I am trying to modify the "Semantic Segmentation Using Deep Learning" example found at the following link so that it can be trained using multiple GPU's:

https://www.mathworks.com/help/vision/examples/semantic-segmentation-using-deep-learning.html

If I try to have the the example use multiple GPU's by adding 'ExecutionEnvironment','multi-gpu' to the "trainingOptions" call, I get the following error when I try to train the network:

Error using trainNetwork (line 140)

Parallel training of DAG networks is not supported. 'ExecutionEnvironment' value in trainingOptions function

must be 'auto', 'gpu', or 'cpu'.

How can I use multiple GPU's on DAG networks?

Best Answer

This functionality is supported as of MATLAB R2018a. Unfortunately, there is no workaround for previous versions.

Related Solutions

MATLAB: How to generate code for the deep neural network with ScalingLayer

The workaround is to substitute the 'ScalingLayer' with an 'nnet.onnx.layer.ElementwiseAffineLayer'. However, this would involve using a attached p-coded script ('getNetworkFromRepresentation.p') to extract the network from the 'rlLayerRepresentation' object. The steps are outlined in the script below:

%%load agent MAT file
load agent
%%get actor representation (contains scalingLayer) from saved agent
actor = getActor(saved_agent);
%%get the new network with codegen-supported layer
actorNetwork = getNetworkFromRepresentation(actor);
newActorNetwork = actorNetwork;
% replace any scalingLayer with nnet.onnx.layer.ElementwiseAffineLayer. The
% name, scale, bias values of old scalingLayer are transfered to the new
% layer
for ct = 1:numel(newActorNetwork.Layers)
    if isa(newActorNetwork.Layers(ct),'rl.layer.ScalingLayer')
        newLayer = nnet.onnx.layer.ElementwiseAffineLayer(...
            newActorNetwork.Layers(ct).Name,...
            newActorNetwork.Layers(ct).Scale,...
            newActorNetwork.Layers(ct).Bias);
        newActorNetwork = replaceLayer(newActorNetwork,newActorNetwork.Layers(ct).Name,newLayer);
    end
end
%%recreate the representation with net network
actionLayerName      = newActorNetwork.Layers(end).Name;
observationLayerName = newActorNetwork.Layers(1).Name;
actionInfo      = actor.ActionInfo;
observationInfo = actor.ObservationInfo;
newActor = rlRepresentation(newActorNetwork,observationInfo,actionInfo,...
    'Observation',observationLayerName,'Action',actionLayerName,actor.Options);
saved_agent = setActor(saved_agent,newActor);
%%codegen
generatePolicyFunction(saved_agent)
cfg = coder.gpuConfig('mex');
cfg.TargetLang = 'C++';
cfg.GpuConfig.Enabled = true;
cfg.DeepLearningConfig = coder.DeepLearningConfig('cudnn');
argstr = '{ones(4,1)}';
codegen('-config','cfg','evaluatePolicy','-args',argstr,'-report');

Notes:

You are free to use the 'getNetworkFromRepresentation.p' script as you wish.
You will need to install the 'Deep Learning Toolbox Converter for ONNX Model Format support package'.

MATLAB: How to specify the number of gpu to train a deep learning model.

There area multiple ways to set gpu prefernces for your training, one of them is using gpuDevice function. Look at the documentation you will get clear idea on how to set different gpu's at the time of training. Although let me give a small piece of code for you to understand the function.

% Set the execution environment to use GPU.
executionEnvironment = "gpu";
% set 3rd gpu from the GPU pool list.
gpuDevice(3);

Also check out this link to get better idea on how to use multiple GPU at the same time. you can also refer to this.

Hope this helps you.

Best Answer

Related Solutions

MATLAB: How to generate code for the deep neural network with ScalingLayer

MATLAB: How to specify the number of gpu to train a deep learning model.

Related Question