how to implement VMD-GRU for timeseries forecasing?

Question

I'm building VMD-GRU model to forecast crude oil price. I have multivariate time series with two columns(crude price, and sentimental index). Each column containes daily observations for ten years. My code structure as follows:

Class GRUModel(nn.Module):
   def __init__(self):
     ....
   def forwaed: 
     ....


def vmd-decomposition(data):
  return trend_price_imdfs, season_price_imdfs, trend_sentiment_imdfs,season_sentiment_imdfs

def prepare_dataset(dataset): 
   return data_X, data_y

the main function:

df= pd.read_csv(file)
imdfs = vmd-decomposition(df)
dataset = prepare_dataset(df) #should I pass the orginal dataset or imdfs
preds = train(model,dataset)

I'm not sure what is the right way to implement this idea(VMD-GRU). Should I pass the four obtained imdfs to the model alongside with orginal data? it will be 6 columns.

or the decomposition should be like a layer in some where?