The NVIDIA® RAPIDS suite of open-source software libraries enable end-to-end data science pipelines to run entirely on GPUs, offering massive parallelism, high memory bandwidth and high speed interconnect through user-friendly PyData APIs. Coupled with Dask, a Python library for distributed computing, RAPIDS libraries can achieve even higher performance by scaling out across multiple GPUs and machines. And now, with the introduction of Dask-SQL, SQL users can take full advantage of this software stack to accelerate their data science workloads. This talk will present an overview of RAPIDS-Dask and introduce Dask-SQL, a developing SQL interface for the RAPIDS-Dask ecosystem which can leverage both CPU and GPUs.
21 авг 2024