adding one more dimension to gluonts deepVAR gives a infer_shape error

[ShirleyMgit]

I have trained gluonts deepAR and deepVAR and I would like to get the predicted distribution parameters without sampling, is it possibe?

[mbohlkeschneider]

Hi @ShirleyMgit,

Thank you for your interest in using GluonTS. I'm not sure I fully understand the question. To clarify, the distribution parameters are neural network outputs, so you can only get them but unrolling the network over some input data. You can check this script on how it works in general, but you will need to invoke predict (and basically sample) still to do the forward pass to compute the distribution parameters.

Hope that helps!