rerun cell with datafame make memory leak

Reproduce the reported memory growth in a Jupyter kernel using the supplied Python dataframe and repeated df_temp.head() cells, then compare it with the versions that call copy(deep=True). Investigate the kernel's handling of displayed dataframe output; done means repeated execution no longer causes the reported retained-memory increase, with a regression test for the reproduction.

env

use vscode remote-ssh connect to debian server and use the jupyter

debian12
vscode=Version: 1.99.0
python=3.10
ipykernal=6.29.5
pandas=2.2.3

replay

create a big dataframe

import pandas as pd
import numpy as np

np.random.seed(0)
num_rows = 10000000  
num_cols = 10  
# create the random number
data = np.random.randint(0, 100, size=(num_rows, num_cols))
df = pd.DataFrame(data, columns=[f'col_{i}' for i in range(num_cols)])

monitor the memory use

%load_ext ipython_memory_usage
%imu_start

when rerun the cell below ,momery will used increasing

df_temp = df.copy(deep=True)
df_temp.head()

in my computer,the result like

[Out] In [4] used 763.6 MiB RAM in 0.55s (system mean cpu 40%, single max cpu 100%), peaked 0.0 MiB above final usage, current RAM usage now 1632.0 MiB

[Out] In [5] used 763.0 MiB RAM in 0.55s (system mean cpu 19%, single max cpu 100%), peaked 0.0 MiB above final usage, current RAM usage now 2394.9 MiB

[Out] In [6] used 763.1 MiB RAM in 0.55s (system mean cpu 25%, single max cpu 100%), peaked 0.0 MiB above final usage, current RAM usage now 3158.0 MiB

try to solve

After many tries , I find the main reason is df_temp.head() ;
If rerun df_temp = df.copy(deep=True) ,the used memory doesn't increase;

If change the code to

df_snapshot_test = df_snapshot.copy(deep=True)
df_snapshot_test.head().copy(deep=True)

rerun will not increase memory too;

I have try

import gc
gc.collect()  

or

from IPython.display 
import clear_output

both can't free the memory

another similar issuse