Experiment Management

Data science and machine learning are iterative processes that require a large number of attempts to reach a certain level of a metric. Experimentation is part of the development of data features, hyperspace exploration, deep learning optimization, etc.

Some of DVC's base features already help you codify and analyze experiments. Parameters are values in a structured text file, which you can tweak and use in your code. On the other end, metrics (and plots) let you define, visualize, and compare quantitative measures of your results.

Experimentation in DVC

New in DVC 2.0 (see dvc version)

DVC experiment management features conform a comprehensive framework to organize, execute, manage, and share ML experiments. They support these main approaches:

• Compare parameters and metrics of existing project versions (for example Git branches) against each other or against new results in the workspace. See for example dvc exp diff.

• Run and capture multiple experiments (derived from any project version as baseline) without polluting your Git history. DVC tracks them for you, letting you compare and share them. 📖 More info in the Experiments Overview.

  New! DVC Experiments can be used directly from the VS Code IDE or online with Iterative Studio, the web UI that integrates all of our data science tools.

• Generate checkpoints at runtime to keep track of the internal progress of deeper experiments. DVC captures live metrics, which you can manage in batches.

👨‍💻 See Get Started: Experiments for a hands-on introduction to DVC experiments.

Organization patterns

It's up to you to decide how to organize completed experiments. These are the main alternatives:

• Git tags and branches - use the repo's "time dimension" to distribute your experiments. This makes the most sense for experiments that build on each other. Git-based experiment structures are especially helpful along with Git history exploration tools like GitHub.

• Directories - the project's "space dimension" can be structured with directories (folders) to organize experiments. Useful when you want to see all your experiments at the same time (without switching versions) by just exploring the file system.

• Hybrid - combining an intuitive directory structure with a good repo branching strategy tends to be the best option for complex projects. Completely independent experiments live in separate directories (and can be generated with foreach stages, for example), while their progress can be found in different branches.

• Labels - in general, you can record experiments in a separate system and structure them using custom labeling. This is typical in dedicated experiment tracking tools. A possible problem with this approach is that it's easy to lose the connection between your project history and the experiments logged.

DVC takes care of arranging dvc exp experiments and the data cache under the hood so there's no need to decide on the above until your experiments are made persistent.

Run Cache: Automatic Log of Stage Runs

Every time you reproduce a pipeline with DVC, it logs the unique signature of each stage run (in .dvc/cache/runs by default). If it never happened before, its command(s) are executed normally. Every subsequent time a stage runs under the same conditions, the previous results can be restored instantly, without wasting time or computing resources.

✅ This built-in feature is called run-cache and it can dramatically improve performance. It's enabled out-of-the-box (can be disabled), which means DVC is already saving all of your tests and experiments behind the scene. But there's no easy way to explore it.