Runtime configuration¶
Through the notebook settings menu, you can configure how and when marimo runs cells.
On startup¶
By default, marimo notebooks run automatically on startup; just how the command
python main.py
executes a script,
marimo edit notebook.py
executes the notebook.
Disable this behavior by unchecking “Autorun on startup”.
When sharing a notebook as an app with marimo run, this setting has
no effect.
On cell change¶
By default, when a cell is run or a UI element is interacted with, marimo
automatically runs cells that reference any of its variables. You can disable
automatic execution of cell’s descendants in the notebook settings menu by
setting "On cell change" to "lazy".
When the runtime is lazy, running a cell marks affected cells as stale but doesn’t automatically run them. Lazy evaluation means cells are only run when their outputs are needed. If you run a cell that has stale ancestors, those ancestors will also run to make sure your cell doesn’t use stale inputs. You can always click the notebook run button or use the keyboard shortcut to run all stale cells.
When should I use lazy evaluation? Choosing the lazy runtime can be helpful when working on notebooks with expensive cells.
Tip: speed up expensive notebooks with marimo’s smart caching
In addition to runtime configuration, marimo also provides opt-in caching to help you work with expensive or side-effectful notebooks. marimo’s can cache expensive functions in memory and expensive blocks of code to disk, letting you skip entire sections of your code and automatically loading variables in memory on notebook startup. Read our caching guide to learn more.
When sharing a notebook as an app with marimo run, this setting has
no effect.
On module change¶
When module autoreloading is enabled, marimo automatically runs cells when you edit Python files. Based on static analysis, the reloader only runs cells affected by your edits. The reloader is recursive, meaning that marimo tracks modifications for modules imported by your notebook’s imported modules too.
Why autoreload?
Autoreloading enables a workflow that many developers find productive: develop complex logic in Python modules, and use the marimo notebook as a DAG or main script that orchestrates your logic.
Autoreloading comes in two types:
autorun: automatically re-runs cells affected by module modification.
lazy: marks cells affected by module modifications as stale, letting you know which cells need to be re-run.