Multivariate time series support for FlowState
Hi, thanks for the great model.
I noticed a discrepancy between the paper, which describes inputs as "c-channel multivariate time series," and the current implementation, which raises an error if the number of input channels is greater than one.
Could you clarify if the model is intended for univariate-only forecasting, or if there are plans to add support for multivariate time series in the future?
Dear @benderlex ,
First of all, my apologies for the very slow response.
Then, thank you for your interest in our model!
It is correct that the model itself can be used for "c-channel multivariate time series forecasting". However, the currently available code showcases only the usage with uni-variate data and raises this error here.
However, we are planning to also release an addition to enabling multi-variate forecasting as well.