machine-learning-foundations / day_01_exercise_intro Goto Github PK

Thank you for taking our course. Completing the following tasks will prepare you for the exercise sessions in the coming weeks. Machine learning on larger scales often requires using central compute clusters, which run on Linux. Consequently, we will use workstations running Ubuntu (a Linux distribution). We highly recommend to use Linux systems instead of Windows.

• Configure GitHub for ssh access. You need to generate a key pair and add the public key to your GitHub account.
  • To generate your key follow the steps here. For Ubuntu the necessary steps are:
    1. Open a terminal by pressing Ctrl+Alt+T.
    2. Execute

    ssh-keygen -t ed25519 -C "[email protected]"

    3. Press Enter twice and remember where you saved the key (this should be prompted/configured in the step above).
  • How to add a SSH key to GitHub is described here: Adding a new SSH key to your GitHub account
• Hit the green Code-button in the upper right corner of this repository. Select the SSH tab and copy the link leading to your repository. This is necessary to clone your github repository onto your local machine.
• Open a terminal by pressing Ctrl+Alt+T.
• Clone this repository by running

  git clone <ssh-link>

  in the terminal and substitute the <ssh-link> with the link you just copied. After pressing Enter your repository will be downloaded into your current working directory.
• Navigate into the downloaded directory by typing cd exercise_intro-yourname. Use ls to list the contents of the folder you are currently working in. If Visual Studio Code is installed correctly you can open it from the terminal by typing code ..

In Vscode, you can now open a rendered version of this readme. Right-click the file and select Open Preview.

To develop and execute our python code, we use a python container software called miniconda. Using miniconda you can create an environment which holds python and all the required software to run the given scripts.

• Navigate to https://docs.conda.io/en/latest/miniconda.html in your favourite browser. The HRZ-Pool computers run Ubuntu Linux. Download the Miniconda3 Linux 64-bit file.

• Open the terminal on your machine by pressing Ctrl+Alt+T. Navigate into the Downloads folder by typing cd Downloads. Before running the installer, set the executable bit by typing chmod +x Miniconda3-latest-Linux-x86_64.sh. Install Miniconda via ./Miniconda3-latest-Linux-x86_64.sh.

• Finally, execute this command source ~/.bashrc. Check if you can see the (base) environment name on the left-hand side of your command line. This means that (mini)conda is installed correctly.

• Open Visual Studio Code (Vscode).
• Click on the extensions tab in Vscode (on the left hand side) or press Ctrl+Shift+X. Install the Python and Remote-SSH extensions. Choose the versions provided by Microsoft.
• Make the Miniconda interpreter your default in Vscode by pressing Ctrl+Shift+P. Type select interpreter and press enter. In the following dialogue, choose the base environment.
• (Optional) For the course, we suggest to install TODO Highlight extension provided by Wayou Liu. This is handy in identifying TODO parts of exercise much easier.

• Open a terminal by pressing Ctrl+Alt+T. Navigate into this directory by typing cd exercise-01-intro-yourgitname. Type

  pip install -r requirements.txt

  to install the Python packages required for this exercise.

Scientific software must provide reproducible results, and automatic testing ensures that our software runs reliably. For example, the recent CrowdStrike incident, which won the 2024 'Most Epic Fail' award highlights the importance of thorough testing.

To prevent similar issues, we strongly recommend testing your code. Let's learn a lesson from the humorous Vincent, " former CrowdStrike employee ", in the image above! We suggest using Nox for test automation.

• To run some of the tests we prepared for you type

  nox -s test

  The python extension provides test integration into Vscode. To use it, click on the lab-flask icon on the left sidebar. When opening it for the first time, it asks you for a configuration. Click the Configure Python Tests button and select pytest in the ensuing prompt. In the next step, Vscode wants to know the location of the test folder. Choose tests. Vscode will now display your tests on the sidebar on the left. Click the play symbol next to the tests folder to run all tests.

• Open src/my_code.py and finish the __init__ function of the Complex class. The idea here is to implement support for complex numbers (see: https://en.wikipedia.org/wiki/Complex_number for more information about complex numbers). Double-check your code by running nox -s test.

• Click on a line number in my_code.py. A red dot appears. Press the debug_test button in the Testing tab, Python will pause, and you can use the build-in Debug console to explore the data at this point.

• Implement and test the add, radius, angle, and multiply functions.

• Run python ./src/julia.py to compute a plot of the Julia set with your Complex class (see: https://en.wikipedia.org/wiki/Julia_set for more information).
• In src/julia.py use plt.plot and plt.imshow to visualize the julia-set. Feel free to play with c to create different sets.

• Professionally written Python code respects coding conventions. Type nox -s format to have nox format your code for you.

• nox can do even more for you! A basic syntax error at the wrong place can cost days of computation time. Type

  nox -s lint

  to check your code for formatting issues and syntax errors.

• Take a look at https://docs.python.org/3/library/typing.html . nox can help you by checking your code for type problems and incorrect function signatures.

  nox -s typing

At the end of the day after you finished all your tasks we want to save the results and upload them to your online GitHub repository. Ideally, all the tests were successful. Follow these steps:

• Open a terminal by pressing Ctrl+Alt+T. Navigate into this directory using the cd command.
• Use

  git status

  to check if there are any files to commit. These are marked red.
• Add all the red files by using

  git add -A

• These files are now staged and can be commited. If you are commiting for the first time you will have to specify your github email address and username by typing

  git config --global user.email "[email protected]"
  git config --global user.name "put_your_username_here"

• Now commit the staged files with

  git commit -m "put_your_commit_message_here"

  and use your own commit message to describe the commit.
• Finally, push everything to GitHub with

  git push

Alternatively, you can use Vscode to commit and push your results.

• For that, go to your code in Vscode and open the source control tab on the left hand side or press Ctrl+Shift+G.
• In the window that opens up enter your individual commit message and press Ctrl+Enter to commit to the main branch.
• Vscode might ask if you want to stage files that haven't been staged yet. Continue with yes. You can also stage files by using the plus sign next to the file.
• Finally, push everything by hitting the Sync Changes button.