Python Environment

This page covers installing OpenVINO from scratch, only if you would like to play with raw OpenVINO API.

1. Install OpenVINO (Python)

Use the official pip package:

python3 -m pip install --upgrade pip
pip install openvino openvino-dev

This installs:

OpenVINO Runtime
Model Optimizer
INT8 calibration tools
CLI utilities

2. Install OS Dependencies (Linux)

Ubuntu 22.04+ (CPU + GPU + NPU)

sudo apt install -y \
    build-essential \
    cmake \
    ocl-icd-opencl-dev \
    intel-opencl-icd \
    intel-level-zero-gpu level-zero \
    clinfo

Check GPU/NPU visibility:

clinfo | grep "Device"

3. Verify Your Installation

Run:

from openvino.runtime import Core
ie = Core()
print(ie.available_devices)

Expected output:

['CPU']
['CPU', 'GPU']
['CPU', 'GPU', 'NPU']   # Meteor Lake AI Boost

4. Meteor Lake NPU Support

NPU requires:

OpenVINO 2024.4+
Linux kernel 6.8+
Intel AI Boost driver (comes with Ubuntu 22.04 HWE or 24.04)
Check NPU availability:

print("NPU" in ie.available_devices)

5. Recommended Device Configurations

CPU (most stable)

compiled = core.compile_model(model, "CPU", {
    "PERFORMANCE_HINT": "THROUGHPUT"
})

iGPU

compiled = core.compile_model(model, "GPU")

NPU (Meteor Lake)

compiled = core.compile_model(model, "NPU")

6. Verify Model Optimizer Availability

mo --help

If MO works, your install is correct.

Next Step

➡️ Model Preparation