diff --git a/README.md b/README.md index 1f722d5..2d3175e 100644 --- a/README.md +++ b/README.md @@ -7,11 +7,9 @@ Python is powering breakthrough exascale scientific discoveries - come learn how supercomputers with it! In this interactive tutorial you'll learn how to write, debug, profile, and optimize high-performance, multi-node GPU applications in Python. You'll learn and master: CuPy for drop-in GPU acceleration of NumPy workflows, Numba for writing custom kernels that match the performance of C++ and Fortran, mpi4py for scaling -across thousands of nodes, DragonHPC for workflow acceleration, and CppJIT for on-demand Python/C++ interoperability -and automatic binding generation, including compiler-based automatic differentiation of C++ with Clad. Everything is -hands-on: short, interactive lectures by expert instructors are paired with guided Jupyter notebooks that introduce +across thousands of nodes, DragonHPC for workflow acceleration with cuPhoton for astronomical image processing, and CppJIT for on-demand Python/C++ interoperability and automatic binding generation, including compiler-based automatic differentiation of C++ with Clad. Everything is hands-on: short, interactive lectures by expert instructors are paired with guided Jupyter notebooks that introduce each concept and have you apply it immediately in bite-sized exercises, among them a mini-app applying several -programming models (CuPy, JAX, cuda.compute, Numba CUDA and CppJIT) to cosmic-web data and a CARTopiaX CAR T-cell +programming models (JAX, cuda.compute, Numba CUDA) to simulated cosmic-web data and a CARTopiaX CAR T-cell therapy example. We'll use a web-based environment that requires no installs or setup - just a laptop and a browser. Whether you're a domain scientist seeking faster turnaround or a software engineer evaluating portable acceleration strategies, you'll leave with a roadmap, skills, and code for bringing scalable Python practices back to your own HPC @@ -48,14 +46,19 @@ helpful but not required. The outline and the notebook list are in The tutorial runs in the container image `ghcr.io/pyhpc/sc2026-tutorial`. [brev/docker-compose.yml](brev/docker-compose.yml) starts JupyterLab on port 8888, opening on the outline, with Nsight Systems on port 8080 and Nsight Compute on -port 8081 as launcher tiles. Brev starts this file. Locally, with Docker and the NVIDIA Container Toolkit installed: +port 8081 as launcher tiles. Brev starts this file; [docs/brev.md](docs/brev.md) is the attendees' guide. Locally, +with Docker and the NVIDIA Container Toolkit installed: ```bash docker compose -f brev/docker-compose.yml up -d docker compose -f brev/docker-compose.yml down -v ``` -## Development + + +
+ +Development ```bash docker compose -f brev/docker-compose.yml -f brev/docker-compose.dev.yml up -d # this checkout mounted in the containers @@ -71,6 +74,9 @@ workflow builds the image on a self-hosted GPU runner set up with `brev/runner-s in it, starts the Compose stack from it and publishes it: a pull request from a branch of this repository gets the tag `pr-`, a push to `main` the tags `git-` and `latest`. + +
+ ## Licences Written material is licensed under CC BY-NC-SA 4.0 ([LICENSE](LICENSE)) and code under Apache-2.0 diff --git a/docs/brev.md b/docs/brev.md new file mode 100644 index 0000000..ed1f5a0 --- /dev/null +++ b/docs/brev.md @@ -0,0 +1,32 @@ +# Running the tutorial on NVIDIA Brev + +[NVIDIA Brev](https://brev.nvidia.com) runs the tutorial on a cloud GPU and opens JupyterLab in your browser. +You need a laptop and a browser. + +## Join the organization + +- Open the invite link from your instructor. +- Log in, or sign up with your email: choose a password, enter the 6-digit code from your email, then create + the NVIDIA Cloud Account with your name. +- The organization name at the top right must read `SC26-Python-HPC`. If it does not, click it and switch. + +## Start the tutorial + +- Click "Launchables" in the top bar, then "My Launchables". +- Click "View Launchable" on `pyHPC-SC26`, then "Deploy Launchable", then "Go to Instance Page". +- Wait for both status indicators to turn green: "Running" after 1 to 3 minutes, "Built" after another + 4 to 10 minutes. +- Click "Open Notebook" when it flashes green. If it does not within a minute, refresh the page. +- JupyterLab opens on the tutorial outline. Its table links every notebook of the day. + +## Work in JupyterLab + +- Exercises have `TODO` cells. The answers are in the `solutions/` folder next to each notebook. +- Run one notebook at a time. Before opening the next, shut down the current kernel + ("Kernel > Shut Down Kernel") to free the GPU memory. +- Nsight Systems and Nsight Compute open from the "nsys" and "ncu" tiles on the JupyterLab launcher. +- Your edits live on the instance. To keep a notebook, right-click it and choose "Download". + +## Afterwards + +Stop or delete the instance from its page in the Brev console.