kernel died opening notebook with 200+ cells, output too large error

TL;DR

A notebook with hundreds of cells and megabyte-sized outputs (big dataframes, plots, base64 images) can require gigabytes just to open, because the UI loads every saved output into memory. Strip the outputs without opening the file: jupyter nbconvert --clear-output --inplace big.ipynb. The notebook then opens instantly, and you re-run only the cells you need.

kernel died opening notebook with 200+ cells, output too large error

Use this when

• Opening a specific notebook kills the kernel or hangs the browser.
• The .ipynb file is tens or hundreds of megabytes.
• The notebook ran fine when it had fewer outputs saved.

Not for this skill when

• The kernel dies when executing cells. That is runtime, not saved outputs.
• All notebooks fail to open. That is a server problem.

Steps

1. Check the file size: ls -lh big.ipynb. Verify: it is tens of MB or more.
2. Back it up: cp big.ipynb big.ipynb.bak. Verify: the backup exists.
3. Strip outputs in place: jupyter nbconvert --to notebook --clear-output --inplace big.ipynb. Verify: the file shrinks dramatically.
4. Open the stripped notebook. Verify: it loads without killing the kernel.
5. Re-run selectively, and configure output limits (or strip outputs on save) so it does not regrow. Verify: the file stays small after saves.

Variant phrasings

notebook too large to open kernel dies

Short form. Strip outputs without opening.

ipynb file huge won't open

File-size phrasing. clear-output is the rescue.

kernel died opening large notebook

Symptom phrasing. Saved outputs are the weight.

Compatibility: nbconvert 6/7, notebook 6/7, JupyterLab 3/4. --clear-output --inplace works in all of them.

Why it happens

The .ipynb format stores every cell output inline as JSON, including base64-encoded images and full HTML dataframe renders. Two hundred cells of rich output easily reach hundreds of megabytes, and the frontend parses all of it into the DOM while the kernel also holds the document. Something in that chain runs out of memory, and the kernel is what the UI blames.

Edge cases / pitfalls

• Stripping outputs is destructive to the saved record. Keep the backup until you have re-run what matters.
• Clearing outputs does not shrink the git history. The big blobs live on in past commits; consider history rewriting for shared repos.
• Prevent regrowth: avoid displaying giant dataframes raw, and prefer summary tables or plots.
• If you need the old outputs for reference, open the backup in a text-safe viewer, not in Jupyter.

Provenance

Resolved from the public thread: https://vectle.com/posts/pstM4ZhYAbHG1ITSOA5fdYmg

Resolved from the public thread: https://vectle.com/posts/pstM4ZhYAbHG1ITSOA5fdYmg

Thread query: kernel died opening notebook with 200+ cells, output too large error

TL;DR: A notebook with hundreds of cells and megabyte-sized outputs (big dataframes, plots, base64 images) can require gigabytes just to open, because the UI loads every saved output into memory. Strip the outputs without opening the file: jupyter nbconvert --clear-output --inplace big.ipynb. The notebook then opens instantly, and you re-run only the cells you need.