download kaggle data and extract to data/ (note: never commit this directory!) https://www.kaggle.com/competitions/machine-learning-in-science-ii-2026/data
install miniconda if conda missing - https://www.anaconda.com/docs/getting-started/miniconda/main
windows (bash) add command: echo 'eval "$(/c/Users/$USERNAME/miniconda3/Scripts/conda.exe shell.bash hook)"' >> ~/.bashrc
conda create -n pi_car python=3.9
conda activate pi_car
pip install -r requirements.txt
prepare_dataset.py - runs label bias analysis and creates a new train file with weightings for even sampling over speed/angle joint distribution
ssh [uni username]@mlis1@nottingham.ac.uk ssh [uni username]@mlis2@nottingham.ac.uk
https://windows.cloud.microsoft/#/devices - VM service for home access
git --version - install with apt if missing
...add ssh key stuff...
git clone git@github.com:lldvdll/PiCar.git
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
bash Miniconda3-latest-Linux-x86_64.sh
source ~/.bashrc
conda create -n pi_car python=3.9
conda activate pi_car
pip install -r PiCar/requirements.txt
- check gpu status
nvidia-smi - pull latest code
git pull - run conda
source ~/.bashrc - activate environment
conda activate pi_car - install requirements
pip install -r PiCar/requirements.txt - run
python src/train_baseline_wandb.py