Skip to content

Windows

Raymond Lo edited this page Nov 18, 2024 · 49 revisions

Skip Steps 1 and 2 if you already installed Python3 and Git on Windows.

1. Install Python and Git

1.1 Install Python

NOTE: ⚠️The version of Python that is available in the Microsoft Store is not recommended.⚠️ It may require installation of additional packages to work well with OpenVINO and the notebooks.

  • Download a Python installer from python.org. Choose Python 3.9, 3.10 or 3.11 and make sure to pick a 64-bit version. For example, this 3.10.11 installer: https://www.python.org/ftp/python/3.10.11/python-3.10.11-amd64.exe
  • Double-click on the installer to run it, and follow the steps in the installer. Check the box to add Python to your PATH, and to install py. At the end of the installer, there is an option to disable the PATH length limit. It is recommended to click this.

1.1 Install Git

  • Download GIT from this link
  • Double click on the installer to run it, and follow the steps in the installer.

2. Install Drivers for GPU, and NPU (AI PC)

We recommend A "Clean Install" of the WHQL Certified GPU driver to ensure the underlying libraries are correctly configured. https://www.intel.com/content/www/us/en/download/785597/834050/intel-arc-iris-xe-graphics-windows.html

Additionally, for AI PC users, please install the latest NPU driver (or last known working Intel® NPU Driver - Windows* 32.0.100.3053) to avoid any potential issues in compiling NPU kernels. https://www.intel.com/content/www/us/en/download/794734/intel-npu-driver-windows.html

3. Install C++ Redistributable (required), FFMPEG (optional)

FFMPEG

4. Install the Notebooks

After installing Python 3 and Git, run each step below using Command Prompt (cmd.exe), not PowerShell. Note: If OpenVINO is installed globally, please do not run any of these commands in a terminal where setupvars.bat is sourced.

5. Create a Virtual Environment

Note: If you already installed openvino-dev and activated the openvino_env environment, you can skip to Step 6. If you use Anaconda, please see the Conda guide.

python -m venv openvino_env

6. Activate the Environment

openvino_env\Scripts\activate

7. Clone the Repository

Note: Using the --depth=1 option for git clone reduces download size.

git clone --depth=1 https://github.com/openvinotoolkit/openvino_notebooks.git
cd openvino_notebooks

8. Install the Packages

This step installs OpenVINO and dependencies like Jupyter Lab. First, upgrade pip to the latest version. Then, install the required dependencies.

python -m pip install --upgrade pip wheel setuptools
pip install -r requirements.txt

NOTE: ⚠️If you have connection issue on HuggingFace in PRC, please set-up the networking environment by following commands before launching notebooks:

pip install -U huggingface_hub
set HF_ENDPOINT=https://hf-mirror.com

For more information, you can visit https://hf-mirror.com/

9. Launch the Notebooks!

To launch a single notebook, like the PyTorch to OpenVINO notebook

jupyter lab notebooks/pytorch-to-openvino/pytorch-to-openvino.ipynb

To launch all notebooks in Jupyter Lab (localhost only)

jupyter lab notebooks

To launch all notebooks available from any host

jupyter lab notebooks --ip 0.0.0.0

In Jupyter Lab, select a notebook from the file browser using the left sidebar. Each notebook is located in a subdirectory within the notebooks directory.

Troubleshooting

  • Notebooks may have conflicting requirements (i.e., dependencies and versioning issues), and the best way to avoid such is to create a customer kernel for notebooks that have complex requirements. You can find how to create custom kernels for jupyter here.

  • If you have fbgemm.dll issue with optimum-cli, you can follow this thread on the potential fixes. Failed to import pytorch fbgemm.dll or one of its dependencies is missing

  • If you have installed multiple versions of Python, use py -3.10 when creating your virtual environment to specify a supported version (in this case 3.10).

  • If you use Anaconda, you may need to add OpenVINO to your Windows PATH. See the wiki/Conda page.

  • If you see an error about needing to install C++, please either install Microsoft Visual C++ Redistributable or use Python 3.7, which does not have this requirement.

  • If you run into any dependencies issues or errors, we recommend reinstalling the entire virtual environment (i.e., delete the openvino_env folder and create a new one). Since packages are cached, it should not take very long to reset everything.