Installation

Neurenix can be installed through multiple methods depending on your environment and requirements. Choose the method that best fits your needs.

Requirements

Before installing Neurenix, ensure your system meets these requirements:

Python: 3.8 or higher (3.8, 3.9, 3.10, 3.11, 3.12, 3.13, 3.14)
Operating System: Linux, macOS, or Windows
Dependencies: NumPy ≥1.24.0, SciPy ≥1.10.0, typing-extensions ≥4.7.0

For hardware acceleration, additional dependencies may be required. See the Hardware-Specific Setup section below.

Installation Methods

Choose Your Installation Method

Select from pip (recommended), conda, or source installation based on your needs

Install Dependencies

The package manager will automatically install required dependencies

Verify Installation

Run a quick test to ensure Neurenix is working correctly

Install with pip (Recommended)

The easiest way to install Neurenix is via pip:

pip install neurenix

For the latest development version:

pip install git+https://github.com/MilesONerd/neurenix.git

Install with conda

If you use Anaconda or Miniconda:

conda install -c conda-forge neurenix

Or create a new environment with Neurenix:

conda create -n neurenix-env python=3.11 neurenix
conda activate neurenix-env

The conda package uses the recipe defined in conda-recipe/meta.yaml and includes all core dependencies (numpy ≥1.24.0, scipy ≥1.10.0, typing-extensions ≥4.7.0).

Install from Source

For development or to get the latest features:

# Clone the repository
git clone https://github.com/MilesONerd/neurenix.git
cd neurenix

# Install build dependencies
pip install setuptools>=42 wheel setuptools-rust>=1.0.0

# Install package in editable mode
pip install -e .

# Install with Rust extensions (requires Rust toolchain)
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
pip install -e .

Building from source requires:

setuptools ≥42
setuptools-rust ≥1.0.0 (for Rust extensions)
Rust toolchain (for the Phynexus engine)
C++ compiler (GCC, Clang, or MSVC)

Optional Dependencies

Install additional features with optional dependency groups:

pip install neurenix[dev]

pip install neurenix[cuda]

pip install neurenix[distributed]

pip install neurenix[agents]

pip install neurenix[huggingface]

pip install neurenix[dev,cuda,distributed,agents,huggingface]

Optional Dependency Details

Package Group	Includes	Use Case
`dev`	pytest, pytest-cov, black, isort, mypy	Development and testing
`cuda`	pycuda≥2022.1	NVIDIA GPU acceleration
`distributed`	ray≥2.0.0	Multi-node training
`agents`	gymnasium≥0.28.0	Reinforcement learning agents
`huggingface`	transformers≥4.0.0, datasets≥2.0.0	Model hub integration

Hardware-Specific Setup

CUDA (NVIDIA GPUs)

For NVIDIA GPU support:

# Install CUDA support
pip install neurenix[cuda]

# Verify CUDA is available
python -c "from neurenix.binding import is_cuda_available; print(is_cuda_available())"

Requires NVIDIA drivers and CUDA toolkit (11.0 or higher) to be installed separately.

ROCm (AMD GPUs)

For AMD GPU support with ROCm:

# Install ROCm support (requires ROCm SDK)
pip install neurenix

# Verify ROCm is available
python -c "from neurenix.binding import is_rocm_available; print(is_rocm_available())"

TPU (Google Cloud)

For Google Cloud TPU support:

# TPU support is included in the base package
pip install neurenix

# Verify TPU is available
python -c "from neurenix.binding import is_tpu_available; print(is_tpu_available())"

ARM Devices

For ARM-based devices (Raspberry Pi, Jetson Nano, etc.):

# Install with ARM optimizations
pip install neurenix

# Verify ARM features
python -c "from neurenix.device import DeviceType; print('ARM supported')"

Neurenix automatically detects ARM architecture features including NEON SIMD and SVE (Scalable Vector Extensions).

WebAssembly

For browser-based execution:

# Install in Pyodide or Emscripten environment
import micropip
await micropip.install('neurenix')

Verify Installation

After installation, verify Neurenix is working correctly:

import neurenix as nx
from neurenix import Tensor, Device, DeviceType

# Print version
print(f"Neurenix version: {nx.__version__}")

# Create a simple tensor
t = Tensor([1, 2, 3, 4])
print(f"Tensor: {t.numpy()}")

# Check available devices
print(f"CPU available: True")
try:
    from neurenix.binding import is_cuda_available
    print(f"CUDA available: {is_cuda_available()}")
except (ImportError, AttributeError):
    print("CUDA available: False")

print("Installation successful!")

Expected output:

Neurenix version: 2.0.1
Tensor: [1 2 3 4]
CPU available: True
CUDA available: True/False
Installation successful!

Command-Line Interface

Neurenix includes a comprehensive CLI with multiple commands:

# View all available commands
neurenix help

# Check hardware capabilities
neurenix hardware

# Initialize a new project
neurenix init

# Train a model
neurenix run config.yaml

# Serve a model
neurenix serve model.onnx

# Export to ONNX
neurenix export model.pt --format onnx

CLI commands are registered in pyproject.toml and include: init, run, save, predict, eval, export, hardware, preprocess, monitor, optimize, dataset, serve, and help.

Troubleshooting

Import Error: No module named ‘neurenix’

Ensure you’ve activated the correct Python environment and installed the package:

pip list | grep neurenix

CUDA Not Available

If CUDA is installed but not detected:

Check NVIDIA drivers: nvidia-smi
Verify CUDA installation: nvcc --version
Reinstall with CUDA support: pip install --force-reinstall neurenix[cuda]

Build Errors from Source

If building from source fails:

Ensure Rust is installed: rustc --version
Update build tools: pip install --upgrade setuptools setuptools-rust wheel
Try building without Rust extensions: export NEURENIX_NO_RUST=1 && pip install -e .

Missing Optional Dependencies

If features are not working:

# Reinstall with all optional dependencies
pip install neurenix[dev,cuda,distributed,agents,huggingface]

Docker Installation

For containerized deployments:

FROM python:3.11-slim

# Install Neurenix
RUN pip install neurenix

# Your application code
COPY . /app
WORKDIR /app

CMD ["python", "train.py"]

Quickstart

Build your first model with Neurenix

API Reference

Explore the complete API documentation

Get Started

Core Concepts

AI Agents

Reinforcement Learning

Advanced Features

Specialized Modules

Hardware Support

Deployment

Installation

Installation

Requirements

Installation Methods

Install with pip (Recommended)

Install with conda

Install from Source

Optional Dependencies

Optional Dependency Details

Hardware-Specific Setup

CUDA (NVIDIA GPUs)

ROCm (AMD GPUs)

TPU (Google Cloud)

ARM Devices

WebAssembly

Verify Installation

Command-Line Interface

Troubleshooting

Import Error: No module named ‘neurenix’

CUDA Not Available

Build Errors from Source

Missing Optional Dependencies

Docker Installation

Quickstart

API Reference

​Installation

​Requirements

​Installation Methods

​Install with pip (Recommended)

​Install with conda

​Install from Source

​Optional Dependencies

​Optional Dependency Details

​Hardware-Specific Setup

​CUDA (NVIDIA GPUs)

​ROCm (AMD GPUs)

​TPU (Google Cloud)

​ARM Devices

​WebAssembly

​Verify Installation

​Command-Line Interface

​Troubleshooting

​Import Error: No module named ‘neurenix’

​CUDA Not Available

​Build Errors from Source

​Missing Optional Dependencies

​Docker Installation

Quickstart

API Reference

Installation

Requirements

Installation Methods

Install with pip (Recommended)

Install with conda

Install from Source

Optional Dependencies

Optional Dependency Details

Hardware-Specific Setup

CUDA (NVIDIA GPUs)

ROCm (AMD GPUs)

TPU (Google Cloud)

ARM Devices

WebAssembly

Verify Installation

Command-Line Interface

Troubleshooting

Import Error: No module named ‘neurenix’

CUDA Not Available

Build Errors from Source

Missing Optional Dependencies

Docker Installation