Code incredibly slow when run in a notebook wrt terminal

[thistlillo]

I would like to understand why the python code I run in a Jupyter Notebook -- in Visual Studio Code connected remotely to a server -- sometimes becomes incredibly slower than in the terminal -- always connected remotely to the same server. Clearly the code runs in the same Python environment.

I have a gigantic xarray, from which I need to select one variable and turn it into a numpy array.

I cannot share the data, so I will only share the Python code:

import numpy as np
import os
from os.path import join
import pygrib
import xarray as xr
from utils import print_dataset_details
input_dir = "/input_folder" 

print("Reading data from", input_dir)


print("SURFACE DATA")
filename = join(input_dir, "SURFACE.grib")
ds_surf = xr.open_dataset(filename, engine='cfgrib', decode_timedelta=True)

print("SST DATA")
sst = ds_surf['sst']
sst = sst.values
print("SST shape:", sst.shape)

When run in a terminal, I get (I share the image):

When run in a Jupyter notebook, the time needed to perform the same operation goes from seconds to more than a minute.

Why? And what can I do to prevent the system from slowing down?

This is not the only code that is affected by this behaviour, VSC+Jupyter sometimes slows down even when run locally and with other scripts/functions.