MATLAB Answers

0

Deep Learning - Generating code for trained sequence data

Asked by Ramakrishnan Raman on 1 Oct 2018
Latest activity Answered by Shounak Mitra on 8 Oct 2018
I have Sequence data and have trained a network using Deep Learning Framework: Example is per below layers = [ ... sequenceInputLayer(featureDimension) lstmLayer(numHiddenUnits,'OutputMode','sequence') fullyConnectedLayer(100) leakyReluLayer fullyConnectedLayer(100) leakyReluLayer fullyConnectedLayer(100) leakyReluLayer fullyConnectedLayer(numResponses) regressionLayer];
Am able to save the trained network and loaded back, per below.
MytrainedNet = net; save MytrainedNet; load MytrainedNet ; YnewPred = predict(MytrainedNet ,myXTrain{1},'MiniBatchSize',1);
However, I need to generate the code function for the trained network.The genFunction does not support this net object. Subsequently, need to be able to generate the C code to be runnable on a CPU (i.e. with no GPU support)

  0 Comments

Sign in to comment.

Products


Release

R2018a

1 Answer

Answer by Shounak Mitra on 8 Oct 2018
 Accepted Answer

Hi Ramakrishnan,
Unfortunately, we do not support code-gen for LSTMs yet. We plan to add that functionality soon.
Thanks Shounak

  0 Comments

Sign in to comment.