seangtkelley/uor-msc-dissertation-xai-african-storms

University of Reading CCAI MSc Dissertation: Using machine learning to predict the intensification and propagation of East African storms # Using machine learning to predict the intensification and propagation of East African storms ## Metadata - **University**: University of Reading - **Master's Program**: MSc Climate Change and Artificial Intelligence (CCAI) 2024-2025 - **Student**: Sean Kelley - **Primary Supervisor**: Eliza Karlowska - **Co-supervisors**: Kieran Hunt, Andy Turner ## Folder Structure For the instructions below, the use of `REPO_ROOT` will refer to the fully qualified directory path where the repository is cloned. - `REPO_ROOT/` - `data`: Data files (see Data Description section for details) - `std`: Standardized ERA5 data files - `processed`: Processed dataset files (output from `src/create_dataset.py`) - `figures`: Figures and table files generated by code - `report`: Latex report files - `src`: Python code and Jupyter notebooks ## Minimum Setup Instructions The following instructions assume you have `python` installed. ### Conda Recommended for complete python environment isolation. 1. Install conda 2. `conda env create -f environment.yml` 3. `conda activate uor-msc-dissertation-xai-african-storms` ### Venv 1. `pip install venv` 2. `python -m venv venv` 3. `source venv/bin/activate` 4. `pip install -r requirements.txt` ## Basic Usage Instructions 1. Adjust configuration in `src/config.py` as needed. 2. Create the dataset: `python src/create_dataset.py` 3. Create exploratory plots: `python src/create_figures.py` 4. Run the experiments: - `./src/run_exps.sh`: runs all experiments sequentially - `python src/run_exp.py --exp_name EXP_NAME`: run specific experiment 5. Evaluate the experimental results: `python src/eval_exps.py` ### Weights & Biases (W&B) setup - Sign up for a free account at wandb.ai. - Add your W&B API key to the environment: - dotenv file: Create a `.env` file in the root directory with the line `WANDB_API_KEY=your_api_key` - Command line: `export WANDB_API_KEY=your_api_key` - Run either training scripts with W&B enabled using the `--wandb_mode online` …