Intern Skills

A Silverpond intership is not just getting coffee! You will be working on active projects that will be incorporated into Silverpond's core codebase.

In order to be productive we have identified a set of core skills we expect any intern to be proficent in when they start.

This repo is to help potenial interns self assess and practice these core skills:

1. Python: Internally we use Python > 3.11. We expect any intern to have coded at least a small project in Python before. Creating a Github account for your code including the link in your application will go a long way towards demonstrating your proficiency in both Python and Github. If you can flex any other programming language skills in this repo that's a bonus!
2. Python environment basics: Typically a Python project runs in it’s own environment with it’s own dependencies specified in a configuration file, like; requirements.txt or pyproject.toml. If you're used to Conda that's ok too.
3. Unix Terminal: Our development machines (The ones with the big GPUS) all run a Linux based OS called NixOS (for what you'll be doing, just think; Ubuntu). You will be expected to access one of these machines remotely over ssh to do some development. We expect you to be able to navigate around, create, move, copy and remove files and directories confidently. See this YouTube video if you need to learn these skills.
4. Git: Source control using Git is a cornerstone of most jobs involving software development. A non-exhaustive list of common git commands you should be familiar with is; clone, pull, push, fetch, merge, checkout, rebase. We also expect you to be using ssh authentication not https. There are lots of great resources online to brush up on your Git skills if you need to.
5. Vim/Neo-VIM: You may use VSCode or some other IDE for your Python development. Many IDEs provide plugins for remote development over ssh. But it is inevitable you’ll find yourself in a Vim editor at some point. Don’t be one of over 1 million developers that have Googled this in past. Learn some vim! Either on YouTube or in the built-in vimtutor that is on most Unix machines.
6. Remove Development: We have mentioned remote development a lot. You will need to have an ssh key pair setup on your machine so we can copy your private key to our development machines to allow you access. This is the same public ssh key you would likely need to add to your github account to enable ssh authentication.
7. GET A PASSWORD MANAGER AND USE IT!

Your Task

Train and evaluate a shape classification model written in Pytorch.

1. Fork and clone this repo
2. Create a python virtual environment and install the dependancies in requirements-cpu.txt
3. Run python generate_classification_dataset.py
4. Open the notebook Fix the bugs and run the training.
5. Fill in the test and show_prediction functions
6. run nbstripout notebooks/shape_classifier.ipynb to clean up you'r notebook's outputs before committing
7. Push your code
8. Create a Pull Request for your solution
9. Create a video explaining your solution
10. Email your Silverpond contact with a link to the PR and the video

What We're Assessing (What you should show in your video submission)

• The way you approach and solve the problem. For example, sketching out the problem, perusing StackOverflow or related API documentation etc.
• Your confidence using common tools Git, Python, virtual envs
• Your ability to communicate what you're thinking when working through problem. This will require you to explain what you're doing during the screen recording. Videos with no explanations will be an automaticp 'No'

What We're NOT Assessing

• Your video production/editing skills! Provided we can see what you're doing and hear your explanations that's great.
• The overally performance of your solution. We're aware that training models can take time (especially on a cpu). So we're not looking for a perfect model, just something demonstrating that the loss is going down and the output is better than random.

Warnings regarding ChatGPT and other LLMs

This task should not simply be asking ChatGPT how to solve the problem, then copy-pasting the solution. ChatGPT is a useful tool but it is not in the spirit of the task. If you don’t show in your video a thorough explanation of how and why you made various decisions then we will have no choice but to assume you’ve just re-typed the ChatGPT response, which is a waste of time for both you and for us.