diff --git a/README.md b/README.md index 28a8fce94aec62983d79b6d100b9da4536b0e823..4eecc49c65bc51685fa39a0a4e84d4d9e2cb6565 100644 --- a/README.md +++ b/README.md @@ -50,10 +50,37 @@ python -m ipykernel install --user --name=virtualenv No you run the **notebooks** with the `virtualenv` kernel. +Important note: double check if you installed torch+cu118 using: +```bash +pip freeze +``` +Sometimes torch instead of torch+cu118 is installed. If so do the following: + +```bash +pip uninstall torch +``` +```bash +pip install torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 --index-url https://download.pytorch.org/whl/cu118 +``` +or check the following website to find the correct commands for using a different OS or conda instead of pip: [Pytorch - Previous Versions](https://pytorch.org/get-started/previous-versions/) ### Content -... +You can use this repository to reproduce our BWDF forecasting resulst based on pre-trained models. Set up an environment as described above and run the following notebooks to reproduce weeks W1 - W4: ++ DMA_reproduce_w1_predictions.ipynb ++ DMA_reproduce_w2_predictions.ipynb: ++ DMA_reproduce_w3_predictions.ipynb ++ DMA_reproduce_w4_predictions.ipynb + + +These notebooks import functions from: + ++ *_data*: contains functions to load flow data, calendar data and weather data ++ *_features*: contains functions to add date features (based on calendar data) and to encode cyclical features (e.g. hours 1 to 24) ++ *_predictions*: contains functions for reproducing the predictions ++ *_training*: creates data_loaders + +then load preprocessed water demand data from the folder *processed_data* and weather data from the folder *weather*. Using the models in the folder *pre-trained_models* to reproduce our forecasting results. ## Funding