---
title: "Create a run"
description: "Start a run of a TIR pipeline version on a CPU or GPU plan, under an experiment, with your own parameter values."
content_type: "how-to"
product: "pipelines"
platform: "TIR"
url: "/tir/pipelines/how-to/create-a-run/"
topic: ["runs"]
relationships: {"requires":["/tir/pipelines/how-to/create-a-pipeline/"]}
---

# Create a run

Start a run of a pipeline version. You choose the pipeline and version, an experiment to keep the
run under, one resource plan for the whole run, and any parameter values.

## Before you begin

You need a role with create permission on the pipeline service and a pipeline with at
least one version.

## Open the form

Select the icon in the **Create Run** column of a pipeline in **Manage Pipelines**. The same icon is
in the **Create Run** column of **Pipeline Versions**, and the **Create Run** button is on
**Pipeline Runs** and **Scheduled Jobs**. When you open the form from a pipeline or version, that
pipeline and version are already selected.

## Fill in the run details

1. In **Run Name**, keep the generated name or type your own, up to 30 characters.
2. In **Pipeline** and **Pipeline Version**, check the selection. To make a new pipeline or version
   right from the form, use the link beside the field. It opens the create form in a side
   panel, and you return to the run form with the new item selected.
3. (Optional) In **Run Description**, add up to 300 characters.
4. In **Experiment**, choose one. An experiment keeps a record of the pipelines and runs started
   under it. To make one, use the link beside **Experiment** to open **Create Experiment**, type a
   **Name** of up to 30 characters, optionally a **Description** of up to 300 characters, and select
   **Create**.

## Choose the plan

Under **Select Computing Resource**, pick a GPU or CPU plan, then its size in **Configuration**: the
vCPUs and RAM, and for a GPU plan the number of GPUs. A plan marked **Inventory not available** or
**Plan not available** is out of stock. Under **Instance Pricing**, pick a price: **On-Demand** is
hourly, and **Show more pricing** has any others. The whole run uses this one plan, and it
is billed by the minute at the plan's hourly price. Use a CPU plan for steps that prepare or move
data and a GPU plan when the work is training. See [Billing and plans for pipeline
runs](/tir/pipelines/concepts/billing-and-plans/).

## Set the parameters

Under **Run Parameters**, an Argo file's parameters start at their default values. Change a value
to use a different one for this run. Select **Add Parameter** to add a **Key** and a **Value**; both
are required.

To repeat the run on a schedule, turn on **Enable Scheduled Run** on this form; see
[Schedule a run](/tir/pipelines/how-to/schedule-a-run/).

## Start it

Select **Create Run**. You return to **Manage Pipelines**, where the run shows in **Recent Runs**
and on the **Pipeline Runs** tab. For help with a run refused for capacity or credit, see
[Troubleshooting Pipelines](/tir/pipelines/troubleshooting/pipelines/).

To follow the run, [watch it and read its
logs](/tir/pipelines/how-to/watch-a-run-and-read-its-logs/).