Skip to content
| Marketplace
Sign in
Visual Studio Code>Programming Languages>mlsys-lab — CUDA, C++ and LLM-systems exercises, auto-gradedNew to Visual Studio Code? Get it now.
mlsys-lab — CUDA, C++ and LLM-systems exercises, auto-graded

mlsys-lab — CUDA, C++ and LLM-systems exercises, auto-graded

mikamika06

| (0) | Free
Graded exercises in CUDA, C++, Python and LLM systems. Write real CUDA kernels with no NVIDIA GPU: a software GPU counts warps, coalescing transactions, shared-memory bank conflicts and divergence. Deterministic gates, identical score on any machine.
Installation
Launch VS Code Quick Open (Ctrl+P), paste the following command, and press enter.
Copied to clipboard
More Info

mlsys-lab — VS Code extension

The workspace UI: a roadmap of the task bank, the statement, an editor and the grader's verdict, side by side.

Run it

npx @vscode/vsce package
code --install-extension mlsys-lab-0.1.0.vsix

Open the repository root (the folder containing src/mlsys/ and tasks/) and run mlsys-lab: Open Workspace.

How grading is dispatched

The extension shells out to the engine; the language comes from the task's meta.json:

meta.native file you edit command
(absent) solve.py python -m mlsys grade <id> --file solve.py --json
"cpp" solve.cpp python -m mlsys.runners.cpp tasks/<id> solve.cpp
"cuda" solve.cu python -m mlsys.runners.cuda tasks/<id> solve.cu

PYTHONPATH is set to the repository's src/, so a checkout works without installing the package.

Settings

  • mlsys.pythonPath — interpreter used to run the grader (default python3)
  • Contact us
  • Jobs
  • Privacy
  • Manage cookies
  • Terms of use
  • Trademarks
© 2026 Microsoft