training deep network accuary goes to 0 ?

hi i train lstm network and receive bad accuracy during the train what parameters i can change, layers to add

i have lstm network

numHiddenUnits = 100;
numClasses = 2;
layers = [ ...
    sequenceInputLayer(inputSize)
    lstmLayer(numHiddenUnits,'OutputMode','sequence')
    fullyConnectedLayer(numClasses)
    softmaxLayer
    classificationLayer];

maxEpochs = 3;
miniBatchSize = 200;%100;

options = trainingOptions('adam', ...
    'ExecutionEnvironment','cpu', ...
    'GradientThreshold',1, ...
    'MaxEpochs',maxEpochs, ...
    'MiniBatchSize',miniBatchSize, ...
    'SequenceLength','longest', ...
    'Shuffle','never', ...
    'Verbose',0, ...
    'Plots','training-progress');