# From Pandas to GPU in seconds Pandas, **but 150x faster**. No code changes, no drama, just raw GPU acceleration with RAPIDS cuDF under the hood. This setup lets you run Jupyter Notebooks inside WSL2, with everything ready to go. And you **don't need to change a single line of code!** --- ## Step-by-step setup (WSL2 + Miniconda + GPU) ### 0. Install WSL2 and Ubuntu (Windows only) First you need to have virtualization enabled on your machine, maybe you already have it, so I would advise you to try, and if you get an error just enable it in your BIOS, look for Virtualization options and enable anything related to them! After that, open PowerShell as **Administrator** and run: ``` wsl --install ``` This will: - Install WSL2 backend - Install Ubuntu (default distro) - Set everything up for first-time use After reboot, open **Ubuntu** from the Start menu. It will ask you to create a username and password -this is your Linux user. If `wsl --install` doesn't work, try: ``` wsl --set-default-version 2 ``` And install Ubuntu manually from the [Microsoft Store](https://aka.ms/wslstore). ### 1. Install Miniconda in WSL2 ``` wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh bash Miniconda3-latest-Linux-x86_64.sh ``` - Accept defaults, type `yes` to initialize conda - Then restart the terminal: ``` exit wsl ``` --- ### 2. Create and activate your Conda env ``` conda create -n jupyter_env python=3.10 -y conda activate jupyter_env ``` --- ### 3. Install Jupyter and core libraries ``` conda install notebook pandas pyarrow fastparquet plotly -y ``` --- ### 4. Add GPU power: install RAPIDS cuDF ``` conda install -c rapidsai -c nvidia -c conda-forge \ cudf=25.04 python=3.10 'cuda-version>=12.0,<=12.8' -y ``` Works out of the box with CUDA 12.x and modern GPUs like the 5090 RTX. --- ### Step 5 - Get the RAPIDS demo notebook Let’s grab the notebooks first and have it ready: ``` mkdir -p notebooks && cd notebooks wget https://raw.githubusercontent.com/rapidsai-community/showcase/refs/heads/main/getting_started_tutorials/cudf_pandas_demo.ipynb wget https://raw.githubusercontent.com/rapidsai-community/showcase/refs/heads/main/getting_started_tutorials/cudf_pandas_opencellid_demo.ipynb ``` This puts the notebooks in a clean `notebooks/` folder so Jupyter will see it right away. --- ### Step 6 - Fire up Jupyter Notebook ``` jupyter notebook --no-browser --ip=0.0.0.0 --port=8888 ``` Then open your browser at: ``` http://localhost:8888 ``` --- ### cuDF Pandas Opencellid demo notebook ![example](https://hackmd.io/_uploads/BJlwq1ogXee.gif) --- ### cuDF Pandas demo notebook ![image](https://hackmd.io/_uploads/HyjjJogQxg.png) --- You’re now running GPU-accelerated pandas inside WSL2, with zero friction. Same code, new 150x speed!