dvc

Why we included this project

Git stops being a reasonable place to keep datasets and model weights pretty quickly, and DVC is the tool that applies versioning to data the same way Git does for code. The repository holds small metadata files while the actual artifacts live in a content-addressed cache, which you can push to S3, SSH, Google Drive, Azure, or other remotes. A fresh clone then pulls the exact data and models a given commit expects. Beyond versioning, DVC defines reproducible pipelines as stages, records metrics, parameters, and plots across runs, and manages experiment queues, all from the command line inside a normal Git workflow. That makes it a natural fit for data scientists and ML engineers who want reproducibility and experiment history without a heavyweight platform, and teams already on Git will find the learning curve gentle whether they work solo or in a small group.