Tony Kim
Jul 04, 2025 21:25
NVIDIA’s RAPIDS suite model 25.06 unveils new options together with GPU Polars streaming, a unified GNN API, and zero-code ML speedups, enhancing Python knowledge science capabilities.
NVIDIA has introduced the most recent model 25.06 of its RAPIDS suite, a set of CUDA-X libraries for Python knowledge science. This launch introduces a number of groundbreaking options designed to boost computational effectivity and knowledge processing capabilities, in line with NVIDIA.
Polars GPU Engine Enhancements
The brand new launch brings important updates to the Polars GPU engine, initially launched in September 2024. One of many key options is the experimental streaming executor, which permits execution on datasets bigger than the accessible VRAM by knowledge partitioning and parallel processing. This improvement is essential for accelerating analytics operations on extraordinarily massive datasets, scaling from tons of of gigabytes to terabytes. Moreover, the replace introduces a shuffle mechanism to facilitate knowledge redistribution between gadgets and assist multi-GPU execution.
One other enhancement consists of assist for rolling aggregations and expanded column manipulation capabilities, that are significantly helpful for time sequence knowledge evaluation. The GPU engine now additionally helps a wider vary of expressions for datetime column manipulation, resembling .strftime()
and .cast_time_unit()
.
Unified API for Graph Neural Networks (GNNs)
The combination of WholeGraph into NVIDIA’s cuGraph-PyG has led to the creation of a Unified API, which accelerates function fetching for GNNs. This API permits customers to seamlessly transition from a single GPU to multi-GPU or multi-node workflows with out modifying their scripts. The acquainted torchrun
command from PyTorch is used to handle processes, facilitating ease of use for PyTorch customers.
Zero-Code Change ML Enhancements
The RAPIDS 25.06 launch expands its zero-code-change acceleration for machine studying by together with assist vector machines (SVMs) within the cuML library. This enables current scikit-learn workflows utilizing SVMs to profit from GPU acceleration with none code modifications. The replace improves compatibility with scikit-learn, enhancing parameter validation and error dealing with.
Further Platform and Compatibility Updates
The discharge additionally consists of upgrades to the RAPIDS Reminiscence Supervisor (RMM), which now helps the hardware-based decompression engine on NVIDIA Blackwell GPUs. This function guarantees efficiency enhancements in IO-intensive workflows. Moreover, the platform now helps Python 3.13, marking the final launch to assist CUDA 11.
Total, the RAPIDS 25.06 launch delivers important developments for knowledge scientists and builders, specializing in enhanced efficiency and ease of use for GPU-accelerated knowledge processing duties.
Picture supply: Shutterstock