Save, export and restore an experiment¶
Saving a notebook writes its code cells, Markdown and current outputs. Editor synchronization only changes the open document. Save explicitly with the toolbar or Ctrl+S / macOS Cmd+S, and confirm it has finished before stopping the environment.
What can be restored¶
| Item | Behavior when you return |
|---|---|
| Saved notebooks, scripts, JSON and HTML files | Remain readable while the environment is retained and its storage is healthy |
| Editor documents and run options saved in a notebook | Restore canvas and configuration without running them |
| Saved cell outputs | Remain visible, but may belong to an earlier version of the code |
| Python variables, in-memory Jobs and complete Results | Require rerunning code or reading your exported files after a kernel restart or environment stop |
| Session history in the workbench Results view | Is not a complete backup; metadata alone does not rebuild in-memory Results after a refresh |
The 结果 (Results) view locates notebook outputs and collects the latest 20 editor job snapshots from the current page session. Check for stale results after changing code or a canvas. The standalone IDE adaptation also writes editor results into notebook cell outputs; save the notebook to persist them.
Keep a directory for each run¶
The export cell in the notebook introduction creates a separate directory for the raw result, package versions and HTML report. Keep the .ipynb or .py source too, along with parameters, units, shots, seed and simulation settings. For numerical comparisons, also retain the full dependency versions and an experiment note.
Download the files through the file browser. The HTML report opens in a browser; JSON supports later analysis. A screenshot alone cannot reproduce the run and usually omits configuration. Before sharing, check saved outputs and files for personal data, access tokens or other restricted content.
Copying your environment URL does not grant another person access. Share experiment files and environment instructions rather than entry tickets or token-bearing URLs.
Notebook, Code and presentation mode¶
Code saves the notebook, then opens code-server through Jupyter's authenticated proxy. The two editors share files on disk, not Python memory. Avoid editing the same file in both at once. When running Python in Code, verify that its interpreter matches the notebook kernel.
演示 (presentation) hides code inputs while keeping text and output visible. Select it again, or return to Notebook, to restore the inputs. It is a display mode, not a separately published website or a source-access restriction.
Check that you can return to your work¶
- Save the notebook and download a result JSON file.
- Close the notebook tab, reopen it through the file browser, and check its source and outputs.
- When no computation needs to continue, save, stop the environment, then start and enter it again.
- Check the files and rerun definition cells before analysis that depends on their variables.
Closing a tab does not pause a computation. Cancellation is cooperative and may not interrupt numerical work immediately. Stopping the environment ends its processes; permanent deletion also removes files. Export anything you need before deleting it.