hip-python

What it is and what it does

hip-python is a low-level binding layer between Python and AMD's HIP (Heterogeneous-compute Interface for Portability) runtime, currently supporting only the AMD GPU backend. It exposes HIP's compute capabilities through Python and Cython interfaces, allowing developers to write GPU-accelerated code in Python without dropping to C++. The package is in pre-alpha status and actively maintained, with wheels available for Python 3.7 through 3.14 on Linux x86_64 platforms.

This is a foundational library intended for developers who need direct, low-level access to HIP's GPU compute functions. It has no runtime dependencies beyond the HIP runtime itself, which must be installed separately on the host system. The permissive MIT license allows use in any project context.

Use it for:

• Prototyping GPU algorithms in Python before optimizing in C++ or when full C++ integration is unnecessary.
• Building Python-based scientific computing libraries that require AMD GPU acceleration.
• Integrating HIP compute into existing Python data science or machine learning workflows.
• Testing and debugging HIP code with Python's interactive environment and tooling.
• Wrapping HIP functionality in higher-level Python frameworks or domain-specific libraries.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Provides low-level Python and Cython bindings for HIP, enabling direct access to AMD GPU compute capabilities from Python code.

Yes, if you have AMD GPU hardware and need Python-level access to HIP compute. The pre-alpha status and lack of runtime dependencies make it low-risk to evaluate. Install only on Linux x86_64; the MIT license and active maintenance are positive signals, but verify that the HIP API surface matches your use case before committing to it in production.

Install

hip-python on PyPI

pip install hip-python

uv add hip-python

poetry add hip-python

Installing hip-python

Before you install

Medium install friction due to platform-specific wheels (x86_64 Linux only). Package is actively maintained with recent commits, though marked pre-alpha and still early in its release cycle.

License in practice

MIT license permits unrestricted use, modification, and distribution with minimal obligations—suitable for commercial and proprietary projects.

Quickstart

pip install hip-python
import hip
# Access HIP runtime functions and GPU compute capabilities

Requires AMD GPU hardware and HIP runtime installed on the system; wheels are Linux x86_64 only.

Verify before relying

• Specific HIP API surface and which HIP runtime versions are supported by this binding.
• Whether Cython bindings allow custom kernel compilation or only wrap existing HIP functions.
• Performance characteristics and overhead compared to direct HIP C++ code.

Package facts

Evidence: hip_python-7.2.2.562.43-cp310-cp310-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl; hip_python-7.2.2.562.43-cp311-cp311-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl; hip_python-7.2.2.562.43-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl; hip_python-7.2.2.562.43-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl; hip_python-7.2.2.562.43-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl; hip_python-7.2.2.562.43-cp39-cp39-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl

Tags

More Libraries packages