Registry / data / nvidia-cufile

nvidia-cufile

library1.18.1.6pypypi✓ verified 24d ago

The `nvidia-cufile` package distributes the NVIDIA GPUDirect Storage (GDS) C libraries (`libcufile`), which enable direct data transfer between storage and GPU memory. It is a foundational component for GPU-accelerated I/O, typically consumed by higher-level Python bindings like `pygds` or other GPU-aware data libraries. As of version 1.17.0.44, it primarily serves as a backend dependency, providing the necessary low-level C components.

data ai-ml

pip install nvidia-cufile

package ↗homepage ↗

INSTALL

IMPORT

SIG · NVIDIA-CUFILE

nvidia-cufile

datapythonv1.18.1.6

Install

1.8s avg

Import

—

Disk

18MB

Pass rate

10/ 10

Env Coverage10 / 10

glibc

3.9–3.13

musl

3.9–3.13

Install & Compatibility

Where this runs

tested against v0.0.0a0 · pip install

no network on importno background threads

Install × environment matrix

Each cell = how many times install + import succeeded across repeated harness runs. Partial = flaky.
glibc = Debian/Ubuntu slim · musl = Alpine Linux

musl

py 3.10–3.910 runs

installs and imports cleanly · install 0.0s · import 0.000s · 17.8MB

glibc

py 3.10–3.910 runs

installs and imports cleanly · install 1.8s · import 0.000s · 22MB

18MB installed

● package 18MB

Code

Verified usage

Verified import paths — ran on the pinned version, not inferred.

No direct Python API

✓ This package distributes C libraries; use `pygds` for Python bindings.

The `nvidia-cufile` package provides `libcufile` shared libraries and headers. It does not expose a direct Python module for user-level imports. Python applications interact with these libraries via dedicated bindings like `pygds`.

The `nvidia-cufile` package primarily provides the underlying `libcufile` libraries. To interact with GPUDirect Storage from Python, you typically use a binding library like `pygds`. This quickstart demonstrates `pygds` initialization, which implicitly relies on `nvidia-cufile` being installed and properly configured on the system.

import pygds # The 'nvidia-cufile' package itself doesn't offer a direct Python API. # It provides the underlying C libraries for GPUDirect Storage. # Python bindings like 'pygds' utilize these libraries. try: # Initialize GPUDirect Storage through its Python bindings pygds.init() print(f"GPUDirect Storage (via pygds) initialized successfully.") print(f"Is GPUDirect Storage enabled: {pygds.is_enabled()}") print(f"Is GPUDirect Storage initialized: {pygds.is_initialized()}") # To actually perform I/O, you would use functions like pygds.open # on a GDS-enabled file system. print("\n`nvidia-cufile` is installed as the backend for `pygds`.") print("For actual GPUDirect Storage file operations, refer to `pygds` documentation.") except ImportError: print("Error: `pygds` not found. Install it with `pip install pygds` " "to use the `nvidia-cufile` backend.") except Exception as e: print(f"An error occurred during pygds initialization: {e}")

Debug

Known issues

gotchaThe `nvidia-cufile` package primarily distributes the low-level C libraries for GPUDirect Storage (`libcufile`). It does not provide a direct Python API for user interaction. Python developers should typically use higher-level binding libraries like `pygds` (`pip install pygds`) to leverage GDS functionality.

fix

Use a Python binding library such as `pygds` to access GPUDirect Storage capabilities.

affects: All

gotchaRequires an NVIDIA GPU with GPUDirect Storage support, compatible NVIDIA drivers, and a CUDA Toolkit installation. Ensure your system meets these hardware and software prerequisites for GDS functionality to work correctly.

fix

Verify that your hardware and software environment, including GPU, drivers, and CUDA Toolkit, are fully compatible with GPUDirect Storage requirements as specified by NVIDIA documentation.

affects: All

gotchaGPUDirect Storage requires specific file systems (e.g., Lustre, BeeGFS, GPFS, WekaIO) and kernel modules to be properly configured and mounted on the host system. Without this specialized storage setup, GDS operations will fail or fall back to slower, non-GDS paths.

fix

Consult NVIDIA GPUDirect Storage documentation to ensure your storage infrastructure is configured with a GDS-enabled file system and that necessary kernel modules are loaded.

affects: All

Upgrade

Version history

1.18.1.6latest on PyPI · released Jun 29, 2026

Audit

Dependencies

No dependency data recorded yet.

Agent activity

12 hits · last 30 days

node

Resources

packagehttps://pypi.org/project/nvidia-cufile/ ↗

homepagehttps://developer.nvidia.com/cuda-zone ↗