Every .ipynb file can be opened in an editor. Everything written there is in plain text (JSON). For each cell which has the "cell_type": "code" there'd be another key-value pair as "execution_count": <number>. As you might have guessed, that is the prompt numbering. Hence, if the notebook contains code which will take time to execute (as was, in my case) this method would be time efficient.
Now, either you can manually change each execution_count or write a simple script to get the numbering right. To check the results just refresh the notebook in the browser without stopping the kernel. And, everything will be as per your needs, even all the variables/loaded data will remain in the environment.
If what you want is to remove the numbers themselves, so that each cell shows In [ ] (instead of something like In [247] which is leftover from some previous incarnation of the kernel), use "Cell" > "All Output" > "Clear" (in Jupyter Notebook 5.4.0) or "Edit" > "Clear All Outputs" (In Jupyter Lab 0.32.1).
This will remove all the numbers, even if you're in the middle of running a notebook. It will not reset the numbering back to 1; e.g. if the last cell you executed was 18, the next will be 19.
If you're using this because you want clarity about which cells you've executed during this run of the kernel and which cells you haven't executed yet, use "Cell" > "All Output" > "Clear" (or "Edit" > "Clear All Outputs") immediately after you start (or restart) the kernel. This can be useful when restarting a kernel, or when opening a saved or duplicated notebook.
This will also remove all outputs from the notebook.
Thanks to user2651084 in a previous comment for this.
I'm a bit too late, but I had the same problem, and since my notebook had cells with execution time up to 5 minutes, I had to wait a long time until Restart & Run All finished.
So I've made a Python script to make this task for me:
import json
file = '/your/notebook/path/Notebook.ipynb'
# Since every notebook is actually a JSON (JavaScript
# Object Notation), then its contents can be represented
# in a dictionary (or a list of dictionaries)
with open(file, encoding='utf-8') as f:
nb = json.load(f)
count = 1
for cell in nb['cells']:
# Markdown cells doesn't have execution count,
# so apply this only to cells that have one
if 'execution_count' in cell:
cell['execution_count'] = count
count += 1
# Not all code cells have output, such as functions
# that return None or simple declarations, so apply
# this only to cells that have some output
try:
for output in cell['outputs']:
if 'execution_count' in output:
output['execution_count'] = cell['execution_count']
except KeyError:
continue
with open(file, 'w+') as f:
json.dump(nb, f, indent=2, ensure_ascii=False)
But be careful with the execution order and the variables in your cells, since applying the script above on your notebook can generate a different output if you run the notebook again. For example, let's suppose your notebook have the following cells with the execution order in square brackets:
In [2]: a = 1
In [1]: a = 2
In [3]: a
Out[3]: 1
If you apply the above script into your notebook, it'll show the following:
In [1]: a = 1
In [2]: a = 2
In [3]: a
Out[3]: 1
But if you run the notebook again, it'll show the following:
In [1]: a = 1
In [2]: a = 2
In [3]: a
Out[3]: 2
This can be a bit confusing for people who are downloading your notebook via GitHub for example, since they can see an output in the repository, but when they run on their machine, the output will be different.
This is useful when you want to reset all variables after a certain point in the notebook. It is going to ask if you are sure that you want to reset. If you want to force reset without asking, then use: