Upload Jupyter notebook checkpoints#
Not actively maintained
We do not currently maintain the Neptune-Jupyter extension. We have plans of rebuilding it in the future.
Existing users can still use the current version, with the following caveats:
- You need to be on version
0.16.xor below of the old Neptune client library called
neptune-client. You cannot use this extension with version
- We deprecated part of the feature that associates notebook checkpoints with runs. As such, the Activate button in the extension toolbar currently serves no purpose.
- (If using JupyterLab) Only works with version
3.xor below of JupyterLab.
Once you've installed the Neptune-Jupyter extension, you can snapshot your code by uploading notebook checkpoints.
- You can later download a checkpoint from Neptune with the Download button in Jupyter.
- You can also download checkpoints as
.ipynbfiles directly from the Neptune app.
To upload the current notebook as a checkpoint:
- In the Jupyter top navigation, click Upload.
- Select a project from the list.
- Optionally enter a checkpoint name and description. This makes it easier to identify and compare checkpoints in Neptune.
- Click Upload checkpoint.
Click the link in the confirmation message to open the notebook in Neptune.
Automatic upload when starting a run#
You can save notebook checkpoints automatically whenever you create a Neptune run in a notebook cell.
To connect your notebook and run, log the notebook to Neptune by clicking Upload in the Jupyter navigation.
After that, whenever you create a run in the notebook with
neptune.init_run(), the checkpoint will be logged automatically.
Missing notebook checkpoints
When you restart the Jupyter kernel, your notebook checkpoints may not be saved to Neptune.
To fix that, reactivate your notebook by clicking Activate in the Jupyter navigation menu.
I changed my Neptune project, but the checkpoints are not appearing
If you want to move the notebook checkpoints to a new project, you need to first create a new notebook upload in the new project.
To go back to the previous project, download a checkpoint from there and continue logging from that point.