libkvikio-cu12
KvikIO - GPUDirect Storage (C++)
Decision gist · record as of 2026-08-14
Yes, if you have an NVIDIA GPU and need to move large arrays between device memory and disk. The medium install friction (CUDA toolkit and cuFile dependency) is justified by the performance gains in GPU-accelerated workloads. Active maintenance, permissive Apache-2.0 license, and zero known vulnerabilities support adoption. Not suitable for CPU-only environments.AI-flagged interpretation of the facts on this page — verify before relying
Before you install
- Requires NVIDIA GPU, CUDA toolkit, and cuFile library (nvidia-cufile-cu12) installed and properly configured.
- Medium install friction due to CUDA toolkit and GPU-specific dependencies (cuda-pathfinder, cuda-toolkit, nvidia-cufile-cu12).
- Active maintenance with recent releases; repository is current and well-maintained.
License · maintenance · safety
Apache-2.0 (permissive) — Apache-2.0 permissive license allows commercial and private use with minimal restrictions; suitable for most deployment scenarios.
last release 2026-08-06 (8 days) · last repo commit 2026-08-14 · 273 stars
0 known vulnerabilities (OSV.dev, 2026-08-14) · 417,460 downloads/mo, #6,817 on PyPI
Alternatives
Verify before relying
pip install libkvikio-cu12
# Requires GPU array library and file I/O operations
# See documentation at https://docs.rapids.ai/api/kvikio/nightly/- Whether rapids-logger is a hard runtime requirement or optional dependency for logging.
- Exact Python version support (classifiers list 3.11–3.14 but requires_python is unspecified).
- Performance gains over standard file I/O in typical workloads and memory configurations.
- Specific usage patterns and API surface beyond what the description excerpt demonstrates.
What it is and what it does
libkvikio-cu12 is a Python wrapper around NVIDIA's cuFile C++ library that accelerates file I/O operations on GPU memory. It enables direct transfers between GPU device memory and storage without staging through host RAM, a capability known as GPUDirect Storage (GDS). The library also works efficiently when GDS is unavailable, transparently handling reads and writes to both host and device memory.
The package is designed for data-intensive GPU workloads—particularly in scientific computing and data analytics—where moving large arrays to and from disk is a bottleneck. It provides a Python API with context managers and non-blocking operations via an internal thread pool, plus a Zarr backend for seamless GPU data serialization. Installation requires cuda-toolkit, cuda-pathfinder, nvidia-cufile-cu12, and rapids-logger.
Use it for
- Accelerate data loading pipelines that move large arrays directly into GPU memory without host RAM staging.
- Implement high-throughput concurrent reads and writes using the internal thread pool for data-intensive workloads.
- Serialize and deserialize GPU arrays to file efficiently using the Zarr backend for checkpoint operations.
- Build GPU-native data processing pipelines where file I/O latency is a critical performance bottleneck.
- Enable seamless host–device memory I/O in scientific computing applications without explicit memory transfers.
Worth the install?
AI-flagged interpretation of the facts on this page. Verify before relying on it.
Yes, if you have an NVIDIA GPU and need to move large arrays between device memory and disk.
The medium install friction (CUDA toolkit and cuFile dependency) is justified by the performance gains in GPU-accelerated workloads. Active maintenance, permissive Apache-2.0 license, and zero known vulnerabilities support adoption. Not suitable for CPU-only environments.
Install
libkvikio-cu12 on PyPI
Before you install
Medium install friction due to CUDA toolkit and GPU-specific dependencies (cuda-pathfinder, cuda-toolkit, nvidia-cufile-cu12). Active maintenance with recent releases; repository is current and well-maintained.
Requires NVIDIA GPU, CUDA toolkit, and cuFile library (nvidia-cufile-cu12) installed and properly configured.
License in practice
Apache-2.0 permissive license allows commercial and private use with minimal restrictions; suitable for most deployment scenarios.
Quickstart
pip install libkvikio-cu12
# Requires GPU array library and file I/O operations
# See documentation at https://docs.rapids.ai/api/kvikio/nightly/Verify before relying
- Whether rapids-logger is a hard runtime requirement or optional dependency for logging.
- Exact Python version support (classifiers list 3.11–3.14 but requires_python is unspecified).
- Performance gains over standard file I/O in typical workloads and memory configurations.
- Specific usage patterns and API surface beyond what the description excerpt demonstrates.
Package facts
| License | Apache-2.0 permissive |
| Python support | Not specified |
| Install friction | Medium. Platform-specific wheel |
| Runtime dependencies | 4 packagescuda-pathfindercuda-toolkitnvidia-cufile-cu12rapids-logger |
| Maintenance | Actively maintained 8 days since the last release |
| Last repo commit | |
| First released | |
| Downloads | 417,460 / month, #6,817 on PyPI 30-day window, as of 2026-08-14 |
| Known vulnerabilities | None known OSV.dev, checked 2026-08-14 |
| Classifiers | Intended Audience :: DevelopersProgramming Language :: PythonProgramming Language :: Python :: 3.11Programming Language :: Python :: 3.12Programming Language :: Python :: 3.13Programming Language :: Python :: 3.14Topic :: DatabaseTopic :: Scientific/Engineering |
Evidence: libkvikio_cu12-26.8.0-py3-none-manylinux_2_28_aarch64.whl; libkvikio_cu12-26.8.0-py3-none-manylinux_2_28_x86_64.whl
Tags
Let your AI agent find packages like this
Example. Real query, live index.
You found this page by searching. An agent finds it by wishing: SkillFed indexes 14,416 PyPI packages by what they can do, searchable in plain language.
wish › “gpu file io cuda”
- libkvikio-cu12libkvikio-cu12 provides Python and C++ bindings to NVIDIA's cuFile…
- cufile-pythonProvides a Python wrapper around the NVIDIA cuFile API to read and…
- instanttensorInstantTensor provides a high-throughput Safetensors loader optimized…
Give your agent the search over MCP, or paste the wish link into any chat.
More Scientific/Engineering packages
NumPy provides an N-dimensional array object and a comprehensive suite of mathematical, linear algebra, Fourier transform, and random number functions for scientific computing in Python.
pandas provides fast, flexible data structures (Series and DataFrame) for loading, cleaning, transforming, and analyzing labeled or relational data in Python.
scipy provides numerical algorithms for mathematics, science, and engineering—including optimization, integration, linear algebra, Fourier transforms, signal and image processing, and ODE solvers—built on numpy arrays.
scikit-learn provides a comprehensive Python library for supervised and unsupervised machine learning, including classification, regression, clustering, dimensionality reduction, and model evaluation tools built on NumPy and SciPy.
Install it if you need to train, evaluate, or deploy supervised or unsupervised learning models.
dill extends Python's pickle module to serialize and deserialize a much wider range of Python objects, including functions, lambdas, classes, and interpreter sessions, to byte streams for storage or network transmission.
Multiprocess is an enhanced fork of Python's standard multiprocessing library that uses dill for better serialization, allowing you to spawn processes with a threading-like API and share complex objects between them.
Install it if you use multiprocessing and encounter pickle serialization limits with lambdas or complex objects.
See also nvidia-cufile · nvidia-cufile-cu12 · cufile-python · pylibcudf-cu12 · libcudf-cu12 · nvidia-libnvcomp-cu12 · cudf-cu12 · cuda-python · dask-cuda · libcuvs-cu12