m5-forecasting

Using Pytorch's LSTM for the m5 forecasting competition (time series forecasting)

Training

To install the prerequisites into a conda environment, run

conda env create -f environment.yml

Create a folder called "data", then add folders called "m5" and "out" under that data directory. Extract the contents of the m5 datasets to the m5 folder (available here). To train, run prepare_m5_dataset then train.py.