Example notebooks (Cookbook)

LabPod Cookbook is a separate, public repository of ready-to-run environments and starter notebooks - the fastest way to go from “I have a workspace” to “I’m running something real,” across a range of research domains, not just deep learning.

Each entry in the cookbook is a small LabPod template bundle (a .tar file you import, same as any other template) plus one or more notebooks or example files. The bundle and the notebook are separate on purpose - the bundle just describes the environment (base image, extra packages) and gets built once; your actual work lives in /work, which is what survives stopping and starting a workspace.

What’s there

Domain	Bundle	What it covers
Deep learning (PyTorch)	pytorch-scientific-ml	GPU basics, neural operators, physics-informed neural networks, U-Net, diffusion models, transformers, graph neural networks, reinforcement learning, multi-GPU training
Deep learning (Hugging Face)	huggingface	transformers pipelines - sentiment classification, text generation
Deep learning (MATLAB)	matlab-deep-learning	Image classification with MATLAB’s Deep Learning Toolbox - requires your own MATLAB license
Parallel programming	parallel-programming	The same problem solved with OpenMP, MPI, and CUDA - shared-memory, distributed-memory, and GPU parallelism side by side
Computational fluid dynamics	openfoam-cfd	OpenFOAM’s own lid-driven-cavity tutorial
Molecular dynamics	gromacs-md	GROMACS, pointing at the community-standard external tutorial
Quantum chemistry	quantum-chemistry	Electronic structure calculations with PySCF
Quantum computing	quantum-computing	Quantum circuit simulation with Qiskit
Cheminformatics	cheminformatics	Molecule parsing and similarity search with RDKit
Materials science	materials-science	Crystal structures and lattice optimization with ASE
Seismology	seismology	Real seismic waveform processing with ObsPy
Bioinformatics	bioinformatics-alignment	Read alignment with bwa and samtools
Statistics (R + Python)	r-statistics	Mixed models, GAMs, survival analysis, forecasting, and moving data between R and Python - with both RStudio Server and JupyterLab in one workspace

Most bundles expose a single app. r-statistics is the exception: from the same workspace you can open RStudio or JupyterLab (with R and Python kernels), over one shared Python install - so reticulate from R and the Python kernel are the same environment rather than two you maintain separately. It is CPU-only by design, since classical statistics is CPU- and memory-bound; its README shows how to add GPU PyTorch if you want it.

Each bundle is deliberately minimal - only the dependencies that specific domain needs, not one giant image with everything installed. Most work entirely offline with synthetic or bundled example data; a few (Hugging Face, bioinformatics reference genomes) need internet access or a license the first time you use them - each bundle’s own README says which.

How to use one

1. Pick a bundle from the table above and open its directory in the labpod-cookbook repo - its README has the exact details for that environment.
2. Download its .tar file (linked from the same README, or directly under dist/ in the repo) and import it in LabPod: Images → My templates → Import.
3. Click Build, then Enable the template, then create a workspace from it.
4. Open a terminal in the workspace and clone the repo to get the notebook(s):

   git clone https://github.com/LabPod/labpod-cookbook /work/labpod-cookbook

5. Open the notebook from /work/labpod-cookbook/<bundle>/ and run it.

See Workspaces & templates for more on what importing and building a template actually does.