Skip to content
Edit on GitHub

How to Add Dependencies or Outputs

To add dependencies or outputs to a stage, edit the dvc.yaml file (by hand or using dvc run with the -f --no-exec flags). dvc repro will execute it and cache the output files when ready.

If the stage has already been executed it and the desired outputs are present in the workspace, you can avoid dvc repro (which can be expensive and is unnecessary) and use dvc commit instead.

Note that both alternatives update dvc.lock too.


We start with an example prepare stage, which has a single dependency and output. To add a missing dependency (data/raw.csv) as well as a missing output (data/validate), we can edit dvc.yaml like this:

     cmd: python src/
+    - data/raw.csv
     - src/
     - data/train
+    - data/validate

We could also use dvc run with -f and --no-exec to add another dependency/output to the stage:

$ dvc run -n prepare \
          -f --no-exec \
          -d src/ \
          -d data/raw.csv \
          -o data/train \
          -o data/validate \
          python src/ data/raw.csv

-f overwrites the stage in dvc.yaml, while --no-exec updates the stage without executing it.

If the data/raw.csv or data/validate files already exist, we can use dvc commit to cache the newly specified outputs (and to update the deps and outs file hashes in dvc.lock):

$ dvc commit

🐛 Found an issue? Let us know! Or fix it:

Edit on GitHub

Have a question? Join our chat, we will help you:

Discord Chat