This repo contains all the material I used while attending Weights & Biases' MLOPs course.

The dataset used for this course is the BDD simple 1k, which is a small subset of the original BDD 100k dataset.
The task to be trained on is a semantic segmentation problem.

Checkout the project on Weights & Biases.

Ensure your gpu driver & cuda are properly setup for pytorch to use it (the name of your device should appear):

nvidia-smi

If you don't have it already, install poetry:

make setup-poetry

Setup the environment:

git clone git@github.com:aliberts/wandb-mlops-course.git
cd wandb-mlops-course
conda create --yes --name wandb python=3.10
conda activate wandb
poetry install

Download the bdd1k dataset:

make dataset

This will download and extract the archive into artifacts/ and then delete the original .zip archive. You can also download it manually here. If you do, you'll need to update the dataset.dir option in src/config.py.

Login to W&B website and get your key to paste it later into your terminal when prompted.
Change the wandb.entity option in the src/config.py to yours. You can also change the wandb.project name if you wish to use a different one.

Make your commands shorter with this alias:

alias py='poetry run python'

poetry run python -m src.01_eda
poetry run python -m src.02_data_split
poetry run python -m src.03_baseline