PYTHON-SETUP

Python Setup

601.465/665 assumes familiarity with Python. This guide describes the recommended environment setup for completing homework assignments. It ensures that you’re using the same package versions as the course staff.

You should use Python 3.12 for homework assignments. We recommend working locally with an editor such as VS Code and a Python virtual environment.

The CS Linux machines are also available if you prefer to work remotely.

Working remotely on the CS Linux machines

The CS Linux machines provide a shared Conda environment, nlp-class, with the packages required for this course.

First, obtain a CS account by following the instructions in Dr. Eisner’s Piazza post. Then, connect over SSH to an available ugrad or grad machine (SSH guide).

After logging in, activate the shared environment:

conda activate nlp-class

You can now run Python on the remote machine.

You can also use VS Code remotely through the Remote - SSH extension.

1. Install Python

Install Python 3.12 from the official Python website. Then open a new terminal and verify that you can run it:

macOS or Linux

python3.12 --version

Windows (PowerShell)

py -3.12 --version

Some assignment helper scripts also require Perl and a Unix-style shell. On macOS or Linux, confirm that Perl is installed by running perl --version. On Windows, we recommend using WSL and following the Linux instructions.

2. Set up VS Code (if you haven’t already)

Install Visual Studio Code and Microsoft’s Python extension.

For an introduction, see VS Code’s Python tutorial.

Open the assignment directory in VS Code with File > Open Folder, then use the integrated terminal for the commands below.

3. Create a virtual environment for the NLP class

Open a terminal with Terminal > New Terminal, then follow these instructions to create a virtual environment named nlp-class. You will use this environment for all NLP assignments throughout the semester.

macOS or Linux

VENVS=~/.venvs    # or another directory name you prefer
mkdir -p $VENVS
python3.12 -m venv $VENVS/nlp-class

On Debian or Ubuntu (including WSL), if you’re using the system’s Python rather than one from python.org, the last command may fail with a message that ensurepip is not available. Fix this by running sudo apt install python3.12-venv, then rerun the command.

Windows (PowerShell)

$VENVS = ~/venvs    # or another directory name you prefer
New-Item -ItemType Directory -Force $VENVS
py -3.12 -m venv $VENVS/nlp-class

Optional: For larger or longer-lived Python projects, ⁠uv is a fast, modern alternative for managing Python versions, virtual environments, and dependencies.

4. Tell VS Code to use the environment for your homework

Download the homework materials to a local directory. Open that directory in VS Code and then select the nlp-class environment.

  1. Open the Command Palette (Cmd/Ctrl+Shift+P).
  2. Choose Python: Select Interpreter.
  3. Select the Python interpreter inside $VENVS/nlp-class. If it does not appear in the list, choose Enter interpreter path… and browse to $VENVS/nlp-class/bin/python on macOS or Linux, or $VENVS/nlp-class/Scripts/python.exe on Windows.

VS Code will then associate the nlp-class environment with this project, using it for all of the project’s Python scripts, notebooks, tests, and tools.

5. Tell your shell to use the environment

VS Code will automatically activate nlp-class when you open a new integrated terminal. If you want to run Python from a terminal outside VS Code, or if automatic activation does not occur, you need to first activate nlp-class within that particular shell session.

macOS or Linux

source $VENVS/nlp-class/bin/activate

Windows (PowerShell)

$VENVS/nlp-class/Scripts/Activate.ps1

PowerShell may initially block activation scripts. See the Python venv documentation for the recommended execution policy setting.

6. Populate the nlp-class environment

So far, nlp-class is an empty environment: no special packages will be available. So let’s add the packages required for the assignment. With nlp-class activated, run

python -m pip install --upgrade pip
python -m pip install --upgrade -r requirements.txt --extra-index-url https://download.pytorch.org/whl/cpu

where requirements.txt was provided in the homework directory. For your convenience, we are trying to use the same requirements.txt for the entire course, so you’ll probably only have to do this for HW1 and never again.

If you have an NVIDIA GPU (on Linux or Windows), then change cpu at the end of the URL to cu130, and you’ll get the full version of PyTorch that will use your GPU. This will add a few GB of NVIDIA libraries, though. On a Mac, no change is needed: Apple Silicon GPUs are supported by the standard version.

Note: If your solution needs any additional packages that were not in the starter code, please add them to requirements.txt yourself (rerunning the above commands) and include that file in your submitted code. Then the autograder will be able to find the additional packages and everything will run.

7. Run your code

Each time you work on an assignment:

  1. Open the assignment’s directory in VS Code. VS Code will remember that you were using the nlp-class environment.
  2. Open a new terminal. VS Code should automatically activate nlp-class. Then you can run your code:
python your_program.py