Nvidia has released a new mathematical Python library specialized for Cuda-X. It provides immediate, Python-like access to the core mathematical operations of Cuda-X without using additional C/C++ libraries. It is used to develop hardware-accelerated applications, libraries, frameworks, or deep learning compilers.
Advertisement
Supported by mathematical functions nvmath-python In particular linear algebra and n-dimensional discrete Fourier transforms. Nvidia promises very good performance for these calculations, which should come close to native C libraries. Overhead is reduced and allows, for example, the combination of multiple cores (kernel fusion) without additional hosts.
HashiCorp: New names and features for Terraform, Packer, and VaultThe library is under the Apache 2 license and provides stateless (function style) or stateful (class style) APIs. It integrates into the standard logger library to be able to understand the details of the calculations. In combination with Python compilers such as Numba, device callbacks can be created that customize the behavior of NvMath-Python.
According to Nvidia, NVMath works seamlessly with the Python ecosystem, with other GPU-oriented packages such as CuPy, PyTorch or RAPIDS, as well as with CPU libraries such as SciPy, scikit-learn or NumPy. A new major version of the latter was also released recently.
More information about nvmath-python can be found here Project WebsiteBut GitHub or in one Demo Video,
(Who)
Product Activists: The AARRR Model — How Pirate Metrics Help Product Owners
