0. Environment Setup

To run deep learning code on your local computer, one needs to install different libraries. In this tutorial, we will see how to set up a deep learning environment.

Operating systems

In order to run the code on GPU, install the latest CUDA Driver for your native operating system.

Windows

If you’re a Windows user, install The Windows Subsystem for Linux (WSL).

wsl --install

Follow the instructions and reboot your system.

After that, you simply open the WLS terminal (the penguin icon). The first time you have to update all packages:

sudo apt update && upgrade

Python is already installed in WSL.

libgl error

To use some of the computer vision libraries in WSL you need to install libgl:

sudo apt install --yes libgl1-mesa-dev

CUDA error

If your notebooks crash when calling the cuda() function and get this error: Could not load library libcudnn_cnn_infer.so.8. The solution is to add the library to .bashrc file.

Go to your file browser, navigate to Linux/Ubuntu/home/<user_name>, open the .bashrcfile and add this line to the end of the file:

export LD_LIBRARY_PATH=/usr/lib/wsl/lib:$LD_LIBRARY_PATH

Restart all Ubuntu terminals and the issue should be resolved.

Linux and macOS

There are no extra preparations you need to do.

---

Virtual environment

A virtual environment is a modular solution to install an independent set of packages for different purposes. A “virtual environment” is essentially a single folder that does not affect the whole system and can easily be deleted.

To install a virtual environment with pip, you should follow the official installation guide.

Pip is a package-management system written in Python and is used to install and manage software packages. To install on your system:

sudo apt install python3-pip

Next, we install the virtual environment:

python3 -m pip install --user venv

Creation

To create a virtual environment, you use the command python3 -m venv <path_dir> where <path_dir> is the directory to install the virtual environment. For instance to create a virtual environment with the name “mbb” in the home directory:

python3 -m venv ~/mbb

Activation

To access this environment, you should activate it in the terminal:

source ~/mbb/bin/activate

If activated successfully, you see on the left side of your terminal the name (mbb).

Important Note

You must always remember to activate the virtual environment, for example, before installing any package with pip. Otherwise, it’s installing it in another environment.

Keep in mind that you can activate the same environment in multiple terminals. This is handy for installing packages in one terminal and using the environment in another terminal.

A virtual environment can be deactivated by command:

deactivate

Installing packages

Installing packages with pip is very easy. For instance, to install jupyterlab:

pip install jupyterlab notebook

To run the tutorial codes, we will use the PyTorch framework that can be installed:

pip install torch torchvision torchaudio

ImportError: No module named X

Importantly, if you get an ImportError: No module named X in a Python script. The solution is often easy and correct pip installation commands can be found by googling.

For instance ImportError: No module named cv2 can be resolved by installing opencv-python:

pip install opencv-python