NVIDIANVIDIAOfficialยท7 Agent Skills Included

cudf-spark

Convert Spark UDFs to GPU with tests, benchmarks, and optimization

Converts Apache Spark CPU UDFs into GPU-accelerated versions using cuDF, native CUDA, or Spark SQL. Generates unit tests, CPU-vs-GPU comparison tests, and coverage reports to prove the conversion is correct. Benchmarks and profiles GPU performance, then iteratively optimizes the implementation with measurable results. Includes a review skill that audits conversions for hardcoded test values and hidden CPU fallbacks.
npx skills add NVIDIA/cudf-spark --all -g -y
Available:

Gives AI coding agents the repository's safety rules, GPU resource management patterns, build commands, and shim-layer conventions so they modify the plugin correctly without bypassing project policies.

All Skills in This Repository (7)

Pure Emerald Level Indicators

Frequently Asked Questions

FAQPage Schema
How to install cudf-spark skills?โ–ผ

Run `npx skills add NVIDIA/cudf-spark --all -g -y` in your terminal to install all UDF conversion skills globally.

How to run a Spark UDF on the GPU?โ–ผ

The skills guide your agent through a three-step workflow: generate a unit test, convert the UDF to a cuDF, CUDA, or SQL implementation, then benchmark CPU against GPU.

How do I know my GPU UDF matches the CPU version?โ–ผ

The generated comparison test runs both implementations on the same data and directly compares outputs, while the judge skill audits for hardcoded values and hidden CPU fallbacks.

Can I optimize a slow GPU UDF automatically?โ–ผ

Yes. The udf-optimize-cudf skill profiles your UDF with nsys, applies one targeted change at a time, and keeps only changes that measurably improve benchmark results.

Do these skills work with Claude Code and other AI agents?โ–ผ

Yes. All skills follow the standard SKILL.md format and run in Claude Code, Cursor, and other compatible AI coding agents.

Related Repositories in Software Engineering

View All in Software Engineeringโ†’