NVIDIA’s RAPIDS 25.06 Release Enhances GPU Data Processing for Python
NVIDIA has unveiled version 25.06 of its RAPIDS suite, introducing GPU Polars streaming and a unified GNN API to accelerate Python-based data science workflows. The update enables processing of datasets exceeding available VRAM through partitioning and parallel execution—critical for terabyte-scale analytics.
The Polars GPU engine now supports rolling aggregations and advanced datetime manipulations, streamlining time-series analysis. Multi-GPU execution and memory-efficient shuffling mechanisms further expand large-scale data processing capabilities.