JD-HST-GT / HST-GT

HST-GT: Heterogeneous Spatial-Temporal Graph Transformer for Delivery Time Estimation in Warehouse-Distribution Integration E-Commerce

The Paper has been accepted by CIKM'23!

Run on google colab (recommend)

Our codes, dataset, model and other data are stored in Google Drive （ https://drive.google.com/drive/folders/18MWYE5LteFZLRx-rCHS53ezZTmvvgTU5?usp=sharing ）, and you can train the HST-GT model with Colab.

Train the HST-GT model:

run train.ipynb

with colab, to run the train code successfully, we recommend colab pro+ and choose the gpu option.

If you don't subscribe the Colab Pro+, please use the continue_train.ipynb.

Test the HST-GT model

run test.ipynb

with colab, to run the test code successfully, we recommend choosing the gpu option.

Run on your gpu server

Create the environment(Cuda 11.3)

conda create --name HSTGT --file requirements.txt

Train the HST-GT model:

conda activate HSTGT
python train.py

Test the HST-GT model

conda activate HSTGT
run test.ipynb