cpm-kernels
CPM CUDA kernels
What it is and what it does
cpm-kernels is a library of CUDA kernels designed to accelerate computation in CPM, a large language model framework. It provides GPU-optimized operations for NVIDIA hardware, supporting CUDA versions from 10.1 through 11.5. The package is distributed as a pure Python wheel with no runtime dependencies, making installation straightforward on systems with compatible CUDA toolkits.
However, the package has been abandoned since its latest release on 2022-03-07 with no active maintenance, repository access, or updates. It is classified as alpha-stage software, and there is no documentation or homepage available. Users considering this package should verify whether it remains compatible with their CUDA environment and whether the CPM framework they intend to use still depends on or recommends it.
Use it for:
- Accelerating CPM model inference on NVIDIA GPUs in production systems that have not migrated to newer frameworks.
- Research projects using CPM that require the specific CUDA kernel implementations this package provides.
- Legacy systems running CPM that cannot easily upgrade to alternative or successor libraries.
Worth the install?
AI-flagged interpretation of the facts on this page — verify before relying
Provides CUDA kernels optimized for CPM, enabling GPU-accelerated computation on NVIDIA hardware.
No. The package is abandoned since 2022-03-07 with no active maintenance, repository, or documentation. While it has low install friction and a permissive license, the lack of updates means it is likely incompatible with modern CUDA versions and Python environments. Install only if you are maintaining legacy CPM code that explicitly requires this exact version and your environment is locked to compatible CUDA/Python versions.
Install
cpm-kernels on PyPI
pip
pip install cpm-kernelsuv
uv add cpm-kernelspoetry
poetry add cpm-kernelsInstalling cpm-kernels
Before you install
Low install friction with a pure Python wheel, but the package is abandoned—last release was March 2022 with no active maintenance or repository access available.
License in practice
Apache 2.0 permissive license allows commercial and private use with minimal restrictions, typical for research and ML tooling.
Quickstart
pip install cpm-kernels==1.0.11
import cpm_kernels
Requires NVIDIA CUDA 10.1 or later; package is abandoned and may not be compatible with recent CUDA versions or modern Python environments.
Verify before relying
- Whether the package works with CUDA versions beyond 11.5 (latest listed classifier)
- Compatibility with Python versions released after March 2022
- Whether this package is still required by current CPM implementations or if alternatives exist
Package facts
| License | Apache 2.0 (permissive) |
| Python support | not specified |
| Install friction | low — pure-Python wheel |
| Runtime dependencies | none |
| Maintenance | abandoned — 1,621 days since the last release |
| First released | |
| Downloads | 91,901/month — #13,494 on PyPI (30-day window, as of 2026-08-14) |
| Known vulnerabilities | none known (OSV.dev, checked 2026-08-14) |
Evidence: cpm_kernels-1.0.11-py3-none-any.whl
Keywords: CPM, cuda, AI
Tags
More Python Modules packages
Converts domain names between Unicode and…
permissive · top 100 on PyPI
setuptoolsSetuptools is a Python build backend and…
permissive · top 100 on PyPI
PyYAMLPyYAML parses and emits YAML 1.1 data format,…
permissive · top 100 on PyPI
pydanticPydantic validates Python data structures…
permissive · top 100 on PyPI
annotated-typesProvides reusable metadata objects for use with…
permissive · top 100 on PyPI
typing-inspectionProvides runtime tools to inspect and…
permissive · top 100 on PyPI
nvidia-cudnn-cu11Provides cuDNN runtime libraries for…
unclear · top 5,000 on PyPI
quadrantsQuadrants is a high-performance compiler that…
permissive · top 15,000 on PyPI
sageattentionSageAttention provides quantized attention…
permissive · top 15,000 on PyPI
sgl-deep-gemmsgl-deep-gemm provides optimized GEMM (matrix…
permissive · top 15,000 on PyPI
nvidia-cuda-ccclProvides NVIDIA CUDA C++ Core Compute Libraries…
unclear · top 5,000 on PyPI
slangtorchslangtorch lets you write PyTorch kernels in…
permissive · top 15,000 on PyPI
nvidia-cusolverProvides CUDA solver native runtime libraries…
unclear · top 1,000 on PyPI
nvidia-cublasProvides NVIDIA CUBLAS native runtime libraries…
unclear · top 1,000 on PyPI
nvidia-cuda-runtime-cu11Provides CUDA 11 runtime native libraries for…
unclear · top 5,000 on PyPI
nvidia-cuda-crtProvides NVIDIA CUDA C Runtime libraries for…
unclear · top 5,000 on PyPI