Error using dlfeval when custom training

Hi, when I was doing custom training for the model there was an error of

Error using dlfeval (line 43)

One input argument can have dimension labels only when the other input argument is an unformatted scalar. Use .* for element-wise multiplication.

Error in test_Tranformer (line 81)

[loss,gradients] = dlfeval(@modelGradients, Training_X_1, Training_Y_1, parameters);

Did someone have similar error before and what should I do to fix it?

Answer 1

[Dianxin]

The potential answer could be, do not specific the dimension labels when you do minibatch. Concretely,

mbqTrain = minibatchqueue(cdsTrain,2,...

'MiniBatchSize',minibatch_size,...

'MiniBatchFcn',@preprocessMiniBatch,...

'MiniBatchFormat',{'',''},... % do not specific the dimensional labels or set it carefully

"PartialMiniBatch","discard");

You may get another error by mse function when you cauculate the loss, and if you specific the dimensional labels, say mse(Y, X, "DataFormat", "SSB"), it would be good to go if no addtional errors.