First, we download all the code and some of the pre-trained sample models and dataset from Github.
mkdir JurassicDemo
cd JurassicDemo
git clone https://github.com/IntelSoftware/Jurassic.git
cd JurassicDemo/Jurassic
. prepJurassic.sh
This will download the training dataset. This is required to run the training and more!
cd Jurassic
wget -O data.zip "https://www.dropbox.com/s/ihkbox8vqksb2ii/data.zip" && unzip data.zip
wget -O models.zip "https://www.dropbox.com/s/tj8j6infag4sfcf/models.zip" && unzip models.zip
unzip data.zip
unzip models.zip
NOTE Please complete the two learning units 2 (tabular clustering and classification) and 7 (PyTroch version of resnet18). Feel free to take unit 9 (accelerating inference) using OpenVINO as a homework exercise for the sake of time!
The prepJurassic.sh contains the detailed steps for DevCloud but similar steps for local are needed to be carried out manually.
Install OpenVINO on Local Machine (required for running exercise #9 )
Please see Installation Guide for OpenVINO pre-requisites and requirements. We recommend Python 3.7 on Windows for quick setup. On Linux, please do follow the (Linux GPU Driver Installation)[openvinotoolkit/openvino_notebooks#540] to ensure your system will work with iGPU.
This installation step install OpenVINO 2022.1 with PyTorch, and Jupyter Lab. Please ensure you use the virtual environment to avoid dependencies conflicts.
The prepJurassic.sh has performed these steps already
conda create --clone pytorch --name openvinopytorch
conda activate openvinopytorch
~/.conda/envs/openvinopytorch/bin/pip install --user openvino-dev[pytorch]
python -m ipykernel install --user --name openvinopytorch
pip install -r requirements_openvino.txt
cd Jurassic
python -m venv openvino_env
.\openvino_env\Scripts\activate
python -m pip install -U pip setuptools wheel
pip install -r requirements_openvino.txt
cd Jurassic
python -m venv openvino_env
chmod -R 777 openvino_env/
./openvino_env/bin/activate
python -m pip install -U pip setuptools wheel
pip install -r requirements_openvino.txt
You can start jupyter lab either on DevCloud, or execute the following in the terminal (after installations) locally.
cd Jurassic
jupyter lab .