Neural network clasifing everything into one class (streamlined the problem)

Hi my NN clasifies everything into one class, it does not seem to be learning at all, could you hint me where the problem could be please? I tried everything...No matter the architecture of the neural network it does the same thing, so I think I must use it wrong in some way....If I use patternet the clasification works...

load('workspace2.mat')

%%

lgraph = layerGraph();

tempLayers = [

featureInputLayer(29,"Name","featureinput")

fullyConnectedLayer(32,"Name","fc_1")

fullyConnectedLayer(64,"Name","fc_5")

fullyConnectedLayer(128,"Name","fc_2")

fullyConnectedLayer(2,"Name","fc_3")

classificationLayer("Name","classoutput")];

lgraph = addLayers(lgraph,tempLayers);

plot(lgraph)

analyzeNetwork(lgraph)

%%

categor = categorical(I) ;

X1TrainM = data1;

TTrainM = categor;

dsX1TrainM = arrayDatastore(X1TrainM,IterationDimension=2);

dsTTrainM = arrayDatastore(TTrainM,IterationDimension=2);

options = trainingOptions("sgdm", ...

MaxEpochs=15, ...

InitialLearnRate=0.0000001, ...

Plots="training-progress", ...

Shuffle="every-epoch",...

Verbose=1);

dsTrain = combine(dsX1TrainM,dsTTrainM);

net = trainNetwork(dsTrain,lgraph,options);

YPredicted = classify(net,dsTrain);

plotconfusion(categor',YPredicted)

%This works

t = zeros(2,length(I));

for ind = 1:1:length(I)

t(I(ind),ind) = 1;

end

x = data1;

net = patternnet([64,128,256]);

net.trainParam.max_fail = 20;

net = train(net,x,t);