🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

NVTabular from NVIDIA AI Developer is now built on RAPIDS AI Dask Dask-CuDF backend that simplifies programming while enabling performant preprocessing that scales for #RecSys pipelines and better integration with the RAPIDS ecosystem. nvda.ws/2W7TIli

Unwrap RAPIDS AI release 0.17 from our family to yours: #GPU-Accelerated TreeSHAP in XGBoost. Decimal Types in cuDF. Datetime in cuXFilter. A podcast. And much more! Thanks for your support. Happy New Year! nvda.ws/2W7Laeg

This has been awesome work spearheaded by premsagar and has really helped us to hammer on some of the most complex pieces of code in libcudf.

Dask DataFrame read_parquet's performance for remotely-stored-data has been improved. This will provide faster reads for both the "pyarrow" and "fastparquet" engines, thanks to Rick Zamora!

RAPIDS cuDF uses Filesystem Spec (fsspec) to perform efficient Parquet reads from remote storage. Learn more about the optimizations used in fsspec to make this possible: nvda.ws/3sj7CS5

High-cardinality groupby aggregations on Dask DataFrames are much more performant. Thank you, Rick Zamora! They now use a shuffle-based algorithm, learn more: github.com/dask/dask/pull… /2

Moving between CPU and GPU environments just got a whole lot easier. Find out more in this new blog by Rick Zamora and @quasiben: medium.com/rapids-ai/easy…

Dask DataFrame has loaders for popular formats like CSV and Parquet but historically other formats required you to implement a custom data loader. Now from_map makes custom collection creation both easier and more performant. blog.dask.org/2023/04/12/fro…

RAPIDS cuDF’s pandas accelerator mode is now available in open beta, accelerating #pandas nearly 150x with zero code changes. #DataScience, #Python, #RAPIDS Read the announcement blog: 👇 nvda.ws/469JnqC