From 8928dbb0e94b1559a3d388c562b2d0cdcc4ce2ef Mon Sep 17 00:00:00 2001 From: Emre Sahin Date: Thu, 13 Jan 2022 00:40:01 +0300 Subject: [PATCH] start: Add a parameters hidden section to experiments to merge params/metrics into experiments (#3163) * added a hidden section * Restyled by prettier (#3164) Co-authored-by: Restyled.io * clarification * stages -> experiments * added that * move example out of parens * revised for command argument link to --set-param * replaced "Here is..." with "Example" Co-authored-by: restyled-io[bot] <32688539+restyled-io[bot]@users.noreply.github.com> Co-authored-by: Restyled.io --- content/docs/start/experiments.md | 26 ++++++++++++++++++++++++++ 1 file changed, 26 insertions(+) diff --git a/content/docs/start/experiments.md b/content/docs/start/experiments.md index 35d4d8d5c1..94473306fd 100644 --- a/content/docs/start/experiments.md +++ b/content/docs/start/experiments.md @@ -120,6 +120,32 @@ Experiment results have been applied to your workspace.
+### ℹ️ More information about (Hyper)parameters + +It's pretty common for data science projects to include configuration files that +define adjustable parameters to train a model, adjust model architecture, do +pre-processing, etc. DVC provides a mechanism for experiments to depend on the +specific variables from a file. + +By default, DVC assumes that a parameters file named `params.yaml` is available +in your project. DVC parses this file and creates dependencies to the variables +found in it: `model.conv_units` and `train.epochs`. Example: + +```yaml +train: + epochs: 10 +model: + conv_units: 16 +``` + +When you use `dvc exp run --set-param`, DVC updates the parameters in +`params.yaml` with the values you set in the command line before running the +experiment. + +
+ +
+ ### ⚙️ Run multiple experiments in parallel Instead of running the experiments one-by-one, we can define them to run in a