Closed Vignesh-Sampat closed 2 months ago
Hi @Vignesh-Sampat,
This feature is not supported out of the box, you will need to implement some custom code to run inference only on the encoding part of the models and return the corresponding Tensor
. Since they have different architectures, you will have to have something specific for each MixedCovariatesTorchModel
sub-class.
Out of curiosity, why do you want/need to cluster the embedding of the past covariates series?
I have trained several MixedCovariatesTorchModel families for forecasting models. Is there a way to use only the encoder to extract embeddings from past covariates during inference?