DVC Experiments Overview
ML experiments can be saved with DVC automatically as they're run or manually after they complete. Each experiment creates and tracks a variation of your data science project based on the changes in your workspace.
Experiments preserve a connection to the latest commit in the current branch
(Git HEAD) as their parent or baseline, but do not form part of the regular
Git tree. This prevents bloating your repo with temporary commits and branches.
Experiments are custom Git references (found in .git/refs/exps) with one or
more commits based on HEAD. These commits are hidden and not checked out by
DVC. Note that these are not pushed to Git remotes by default either (see
dvc exp push).
Note that DVC Experiments require a unique name to identify them. DVC will
auto-generate one by default, such as puffy-daks. A custom name can be set
instead, using the --name/-n option of dvc exp run/dvc exp save. These
names can be used to reference experiments in other dvc exp subcommands.
Save experiments
See 👨💻 Get Started: Experiments for a hands-on introduction to DVC experiments.
To save an experiment, you can follow one of these roads:
- If you do not have a DVC pipeline, you can log live results from Python code using DVCLive initialized.
- If you have a DVC pipeline, use
dvc exp runto both run your code pipeline and save experiment results.dvc exp runalso enables advanced features like queuing many experiments at once.
Experiments are saved locally by default but you can share them so that anyone can reproduce your work.
Metrics, plots, parameters
DVC can track and compare parameters, metrics, and
plots data saved in standard structured files like YAML, JSON, and
CSV, and they can be tracked as part of your repo. One way to generate these
parameters, metrics, and plots (and to automatically configure them) is with
DVCLive. You can also manually generate these files and use dvc.yaml
metafiles to specify which files are parameters, metrics, or plots (and to
specify how to visualize plots).
Models and datasets
DVC can track models or datasets as part of your repo, and you can manage those
models in the model registry. One way to log models or other
artifacts is with DVCLive. You can also track them with dvc add
and declare metadata for the model registry in dvc.yaml.
Work with DVC Experiments from a GUI
DVC Experiments can be used directly from the VS Code IDE or online with DVC Studio, the web UI that integrates all of our data science tools.