Registry / ai-ml / nixl-cu12

nixl-cu12

JSON →
library1.4.0pypypi✓ verified 20d ago

NIXL (NVIDIA Inference Xfer Library) is an open-source Python API designed to accelerate point-to-point communications in AI inference frameworks. It provides a unified abstraction layer over various memory types (CPU, GPU) and storage (file, block, object store) through a modular plugin architecture. The `nixl-cu12` package specifically targets CUDA 12 environments. NIXL is actively maintained with frequent releases, with version 1.0.0 being the current stable release.

pip install nixl-cu12
INSTALL
IMPORT
SIG · NIXL-CU12
N
nixl-cu12
ai-mlpythonv1.4.0
Install
79.8s avg
Import
370ms
Disk
5345MB
Pass rate
5/ 10
Env Coverage5 / 10
glibc
3.93.13
musl
3.93.13
Install & Compatibility
Where this runs
tested against v1.4.0 · 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 3.103.910 runs
dependency_conflict
glibc
py 3.103.910 runs
installs and imports cleanly · install 79.8s · import 0.296s · 4915.2MB
5345MB installed
● package 5345MB
Code
Verified usage

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

nixl_agent
from nixl_cu12 import nixl_agent
from nixl import nixl_agent

This quickstart demonstrates how to import the `nixl` library and initialize a NIXL agent. The agent is the core component for managing data transfers within the NIXL framework.

import nixl # Initialize a NIXL agent try: agent = nixl.nixl_agent('my_inference_agent') print(f"NIXL agent '{agent.name}' initialized successfully.") # Further NIXL operations would follow here, e.g., memory registration, transfer requests. except Exception as e: print(f"Error initializing NIXL agent: {e}")
Debug
Known issues
breakingNIXL 1.0.0 introduces significant breaking changes, including the removal of the legacy Multi-Object UCX backend and the complete transition to Device API V2. The previous Device API V1 implementation has been removed.
fix
Review and update your code to utilize the new Device API V2. For UCX backend configurations, adapt to the updated request handling mechanisms. Refer to the official NIXL 1.0.0 release notes and documentation for detailed migration guides.
affects: 1.0.0 and greater
gotchaInstalling both `nixl-cu12` and `nixl-cu13` (or their respective meta-package variants) in the same environment can lead to unexpected behavior. If both are present, `nixl-cu13` will take precedence.
fix
Ensure only the package corresponding to your target CUDA version is installed (`nixl-cu12` for CUDA 12, `nixl-cu13` for CUDA 13). If using the `nixl` meta-package, specify the correct extra, e.g., `pip install "nixl[cu12]"`.
affects: All versions supporting both CUDA 12 and CUDA 13
gotchaNIXL is currently only supported on Linux environments (tested on Ubuntu and Fedora). It does not support macOS or Windows.
fix
Run NIXL in a Linux host, container, or virtual machine environment.
affects: All versions
gotchaErrors such as `NIXL_ERR_BACKEND` often indicate issues with the underlying backend configuration or its availability. This can stem from missing dependencies (like UCX or specific CUDA libraries), incorrect paths, or misconfigurations during agent initialization.
fix
Verify that all required backend libraries (e.g., UCX) are correctly installed and accessible. Check NIXL agent initialization parameters and ensure they match your system's hardware and software configuration. Review NIXL's backend guide for specific setup instructions.
affects: All versions
Upgrade
Version history
1.4.0latest on PyPI · released Aug 14, 2026
Audit
Dependencies
PyTorchrequiredOften used in conjunction with NIXL for AI inference frameworks.
uvoptionalRecommended prerequisite for source builds and managing Python virtual environments.
tomlkitoptionalPrerequisite for source builds.
UCXrequiredHigh-performance communication library backend for NIXL.
Agent activity
36 hits · last 30 days
node
34
OpenAI (training)
1
Resources
nixl-cu12 — pip install nixl-cu12 · libregistry