How to make a model predicting multivariate time series

Hello! I’m new to pytorch and solving problem on Kaggle “Tabular Playground Series - Jan 2022”.
The essence of this task is to figure out whether KaggleMart or KaggleRama should become the official Kaggle outlet. So there are 3 countries, 3 products and 2 markets, that’s why we have to predict num_sold for 18 cases. I know how to make 18 models for predicting 18 cases separately, but it’s very inconvenient. So how can i create 1 model for predicting 18 cases? By the way prediction I perform using LSTM