Is timeshap applicable to the DeepAR model of mxnet?

[togodot]

Hello.
I want to use TimeSHAP to get Feature Importance during multivariate time series predictive analysis with DeepAR model in GluonTS, based on mxnet.

According to the description, I need to use model wrapper, is it applicable to mxnet as well?

I tried TorchModelWrapper, TimeSHAPWrapper in base_wrapper. But I get an error in the parameters property of pytorch.
I was wondering if there is a workaround for this.

Thanks.