hipTensor documentation#
hipTensor is a high-performance C++ HIP (Heterogeneous-Compute Interface for Portability) library for accelerating tensor primitives. It leverages specialized GPU matrix cores on the latest AMD discrete GPUs. The hipTensor API is designed to be portable with the NVIDIA CUDA cuTensor library, letting CUDA users easily migrate to the AMD platform. The hipTensor library is a work in progress (WIP). For more information, see What is hipTensor?
The hipTensor public repository is located at ROCm/hipTensor.
Install
Conceptual
How to
Examples
API reference
To contribute to the documentation, see Contributing to ROCm.
You can find licensing information on the Licensing page.