The post NVIDIA RAPIDS 25.08 Enhances Data Science with New Profiling Tools and Algorithm Support appeared on BitcoinEthereumNews.com. Caroline Bishop Sep 17, 2025 19:45 NVIDIA’s RAPIDS 25.08 release introduces new profiling tools for cuML, updates to the Polars GPU engine, and additional algorithm support, enhancing data science accessibility and scalability. NVIDIA has announced the release of RAPIDS 25.08, an update that continues to advance the capabilities of accelerated data science. This release introduces several new features that enhance the accessibility and scalability of data science processes, according to a blog post by NVIDIA. Introduction of New Profiling Tools The 25.08 version of RAPIDS introduces two new profiling tools aimed at improving the troubleshooting process for cuml.accel code. These tools are designed to help users identify which operations are accelerated on the GPU and which fall back to CPU execution, providing insights into performance bottlenecks in machine learning workflows. The function-level profiler allows users to see the operations executed on both GPU and CPU, and the line-level profiler provides detailed execution information line-by-line. Enhancements to the Polars GPU Engine The Polars GPU engine has been updated to process larger, more complex datasets. The streaming execution mode, previously an experimental feature, is now the default, allowing for the efficient handling of datasets larger than GPU memory. This update supports nearly all operators that are available for in-memory GPU execution, significantly boosting performance and scalability. Moreover, the Polars GPU engine now supports struct data in columns and an expanded set of string operators, keeping more complex data operations on the GPU for improved performance. Additional Algorithm Support in cuML RAPIDS 25.08 also brings new algorithm support to cuML, including Spectral Embedding for dimensionality reduction and manifold learning. The release also adds support for LinearSVC, LinearSVR, and KernelRidge, expanding the range of machine learning algorithms that can be accelerated with zero code changes. Deprecation… The post NVIDIA RAPIDS 25.08 Enhances Data Science with New Profiling Tools and Algorithm Support appeared on BitcoinEthereumNews.com. Caroline Bishop Sep 17, 2025 19:45 NVIDIA’s RAPIDS 25.08 release introduces new profiling tools for cuML, updates to the Polars GPU engine, and additional algorithm support, enhancing data science accessibility and scalability. NVIDIA has announced the release of RAPIDS 25.08, an update that continues to advance the capabilities of accelerated data science. This release introduces several new features that enhance the accessibility and scalability of data science processes, according to a blog post by NVIDIA. Introduction of New Profiling Tools The 25.08 version of RAPIDS introduces two new profiling tools aimed at improving the troubleshooting process for cuml.accel code. These tools are designed to help users identify which operations are accelerated on the GPU and which fall back to CPU execution, providing insights into performance bottlenecks in machine learning workflows. The function-level profiler allows users to see the operations executed on both GPU and CPU, and the line-level profiler provides detailed execution information line-by-line. Enhancements to the Polars GPU Engine The Polars GPU engine has been updated to process larger, more complex datasets. The streaming execution mode, previously an experimental feature, is now the default, allowing for the efficient handling of datasets larger than GPU memory. This update supports nearly all operators that are available for in-memory GPU execution, significantly boosting performance and scalability. Moreover, the Polars GPU engine now supports struct data in columns and an expanded set of string operators, keeping more complex data operations on the GPU for improved performance. Additional Algorithm Support in cuML RAPIDS 25.08 also brings new algorithm support to cuML, including Spectral Embedding for dimensionality reduction and manifold learning. The release also adds support for LinearSVC, LinearSVR, and KernelRidge, expanding the range of machine learning algorithms that can be accelerated with zero code changes. Deprecation…