Repository for ''Contextualizing MLP-Mixers Spatiotemporally for Urban Data Forecast at Scale''

Code structure:

Model:
- nexusqn_model.py
- mlp.py
- embedding.py
Config:
- traffic:
  - default
  - nexusqn_pems08.yaml

main file: run_traffic_benchmark.py

The appendix can be found at urban_forecast_appendix.pdf

More information can be found in our preprint.

Acknowledgement

Our code is built upon the TorchSpatiotemporal repository (https://github.com/TorchSpatiotemporal/tsl).