All skills
accelerated-computing-cudf
Rewrite pandas ETL code to run on NVIDIA GPUs with cuDF, dask-cuDF, or cudf.pandas.
- Part of NVIDIA/skills
- 3.2k installs
- Data · Analytics
When to use it
You have working pandas code and a GPU, but need to move joins, groupby, or CSV/Parquet reads onto the GPU without breaking null or ordering behavior.
For example
You say
Use accelerated-computing-cudf to port my pandas script that reads orders.parquet, joins a customer lookup, and does a groupby sum to cuDF, keeping null handling identical.
You get
- cuDF ETL script
- pandas parity check
- null-handling notes
How it works
- Picks between cudf.pandas, explicit cuDF, and dask-cuDF based on your goal and data size.
- Keeps data on the GPU end-to-end, converting to pandas only at output boundaries.
- Suggests spill or dask-cuDF when the dataset approaches GPU memory limits.
Before installing
Requires a supported NVIDIA GPU and compatible driver, with cuDF, pylibcudf, libcudf, and RMM versions matched to your CUDA major version.
This skill includes 23 scripts your assistant may run on your computer.
More Analytics skills
See all- optimizing-clickhouse-and-hogql-queries148Speeds up slow ClickHouse and HogQL queries and explains how to fix timeouts or memory limits.
- exploring-data69Profiles and analyzes uploaded datasets, including duplicates, joins, drift and time-series checks.
- lance-user-guide7Shows how to read and write Lance datasets and build vector or scalar indexes, including filter and performance tips.
- data_analysis4Runs auditable math and descriptive statistics on data you have already confirmed.
Also in this pack
- cuopt-developer2.8kModifies, builds, tests, and debugs NVIDIA cuOpt source code and prepares contributions.
- cuopt-install2.6kInstalls cuOpt for Python, C, or server via pip, conda, or Docker and verifies it works.
- data-designer2.6kGenerates a synthetic dataset from your description of the data you need.
- dynamo-recipe-runner2.6kSelects, validates, patches, and deploys NVIDIA Dynamo Kubernetes recipes for model serving.