project.qmd

---
title: "Class Project"
---

The goal of this project is to build a training pipeline to preprocess, train, evaluate, and register a machine learning model.

You'll start from the template pipeline that we discussed during the program and make the necessary changes to it. Before making any changes, ensure you can run the pipeline from Session 4 without issues.

The project has three different levels of complexity. Pick the one that you feel most comfortable tackling first.

## Simple complexity
We want to replace the Penguins dataset with a different classification problem. Feel free to use any dataset you like. If you don't have any ideas, here are three options you can choose from:

1. [Iris flowers](https://archive.ics.uci.edu/dataset/53/iris) dataset - This is a multi-class classification problem where you'll predict the flower species given the measurements of iris flowers.
2. [Adult income](https://archive.ics.uci.edu/dataset/2/adult) dataset - This is a binary classification problem where you'll predict whether the income of a person exceeds $50,000/yr based on census data.
3. [Banknote authentication](https://archive.ics.uci.edu/dataset/267/banknote+authentication) dataset - This is a binary classification problem where you'll predict whether a given banknote is authentic given the measures from a photograph.

Start with the pipeline from Session 4 and modify the preprocessing, training, and evaluation scripts to use the new dataset.

## Intermediate complexity
We want to replace TensorFlow with PyTorch in the pipeline we built in Session 4. Everything else will stay the same, except the framework to train the model.

Start with the pipeline from Session 4 and modify the training and evaluation scripts to train and evaluate the model using PyTorch. Notice you'll need to use a [PyTorch estimator](https://sagemaker.readthedocs.io/en/stable/frameworks/pytorch/sagemaker.pytorch.html) to configure the Training Step and a [PyTorch processor](https://sagemaker.readthedocs.io/en/stable/frameworks/pytorch/sagemaker.pytorch.html#pytorch-processor) to configure the evaluation step.

## Advanced complexity
At this stage, we want to combine replacing the Penguins dataset with replacing TensorFlow with PyTorch in the pipeline.

Start with the pipeline from Session 4 and make the necessary changes described in the simple and intermediate complexity sections.