# How to run dvc exp without pulling all dependency graph

**URL:** <https://discuss.dvc.org/t/how-to-run-dvc-exp-without-pulling-all-dependency-graph/2301>\
**Category:** Questions\
**Created:** [November 13, 2024, 5:27pm UTC](https://discuss.dvc.org/t/how-to-run-dvc-exp-without-pulling-all-dependency-graph/2301 "2024-11-13T17:27:04Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![ermolaev](https://yyz1.discourse-cdn.com/flex035/user_avatar/discuss.dvc.org/ermolaev/32/305_2.png) [@ermolaev](https://discuss.dvc.org/u/ermolaev)\
**Post date:** [November 13, 2024, 5:27pm UTC](https://discuss.dvc.org/t/how-to-run-dvc-exp-without-pulling-all-dependency-graph/2301/1 "2024-11-13T17:27:04Z")

</div>

I have the following situation:

- I have huge pipeline with ~25 stages
- I need to run only last 5 of them on ~10 different servers
- I don’t want to download all artifacts from all 25 stages
- I found out that I can run `dvc repro dvc.yaml:train --allow-missing --downstream` and it will work ok if I manually run download of necessary artifacts only for train
- But when I run the same but with `dvc exp` it gives me an error on DVC file that is not downloaded, and I don’t want it to be downloaded, so `--allow-missing` didn’t help here

So, my question is - Is there any way to run experiments without downloading all artifacts? Of corse I assume case when there is no other changes i.e. in code or other. So, all is up to date. I expect from DVC to check that there is no changes and download only those artifacts that are deps for actually running stage. Maybe I missed something or it’s not possible, but for now I can’t get why DVC need to download data that wasn’t changed and is not necessary for current experiment.