IREE
by bCloud LLC
Version 3.12.0 + Free Support on Ubuntu 26.04
IREE (Intermediate Representation Execution Environment) is an enterprise-grade machine learning compiler and runtime platform that enables developers to compile, optimize, and deploy AI models across CPUs, GPUs, and specialized accelerators. Built on MLIR, IREE transforms models from frameworks such as TensorFlow, PyTorch, and ONNX into highly optimized binaries for high-performance, portable AI inference across cloud, edge, and embedded environments.
Features of IREE:
- Compiles machine learning models into efficient native executables for optimal inference performance.
- Supports models from popular frameworks including TensorFlow, PyTorch, and ONNX.
- Leverages MLIR-based compiler technology for advanced graph and hardware-specific optimizations.
- Enables deployment across CPUs, GPUs, and accelerator backends such as CUDA, Vulkan, and ROCm.
- Provides a lightweight runtime suitable for cloud, edge, mobile, and embedded applications.
- Delivers high-performance AI inference with reduced latency and efficient resource utilization.
- Offers portability across diverse hardware architectures through a unified compilation workflow.
- Integrates easily into MLOps, CI/CD, and production AI deployment pipelines.
IREE usage:
$ sudo su
$ cd /opt
$ source ~/iree-env/bin/activate
# Verify installation:
$ pip show iree-base-compiler
$ pip show iree-base-runtime
# Verify compiler and runtime:
$ iree-compile --version
$ which iree-compile
$ which iree-run-module
# Verify Python modules:
$ python -c "import iree.compiler; print('Compiler OK')"
$ python -c "import iree.runtime; print('Runtime OK')"
Disclaimer: IREE is provided under its applicable open-source licensing terms. Users are responsible for ensuring compliance with applicable software licenses, security requirements, export regulations, and organizational policies when compiling, optimizing, deploying, and distributing machine learning models and AI applications using IREE. Hardware-specific performance and supported features may vary depending on the selected backend and deployment environment.