Restart kernel (Ipython Konsole in Spyder 5.1.5) with os._exit(0) does not work when using tensorflow.distribute.MirroredStrategy

Problem Description

Hello,

the problem is described in detail in https://github.com/tensorflow/tensorflow/issues/52135.

I tried it on two Systems Windows Server and Windows 10, both with similar results as described in the link.

I also tried https://docs.spyder-ide.org/current/troubleshooting/common-illnesses.html and https://stackoverflow.com/questions/47267716/spyder-an-error-ocurred-while-starting-the-kernel.

I updated all relevant packages.

I am honest, I did not try reinstalling Anaconda.

But this was not the solution.

I also tried it in jupyter qtconsole with the following errors stats:

2021-10-08 19:44:12.649086: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'nvcuda.dll'; dlerror: nvcuda.dll not found
2021-10-08 19:44:12.649289: W tensorflow/stream_executor/cuda/cuda_driver.cc:269] failed call to cuInit: UNKNOWN ERROR (303)
2021-10-08 19:44:12.656806: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:169] retrieving CUDA diagnostic information for host: RT-Z0M6A
2021-10-08 19:44:12.660749: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:176] hostname: RT-Z0M6A
2021-10-08 19:44:12.661284: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX AVX2
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.

[JupyterQtConsoleApp] KernelRestarter: restarting kernel (1/5), keep random ports

[JupyterQtConsoleApp] WARNING | kernel restarted

[JupyterQtConsoleApp] WARNING | kernel died: 3.0015175342559814

After this the kernel restarted without problems.

The following informations are related to the System with Windows 10:

Versions

• Spyder version: 5.1.5
• Python version: 3.9.7
• Qt version: Jupyter QtConsole 5.1.1
• PyQt version: 5.9.2
• Operating System name/version: Windows 10

Dependencies

# Mandatory:
atomicwrites >=1.2.0          :  1.4.0 (OK)
chardet >=2.0.0               :  3.0.4 (OK)
cloudpickle >=0.5.0           :  1.6.0 (OK)
cookiecutter >=1.6.0          :  1.7.2 (OK)
diff_match_patch >=20181111   :  20200713 (OK)
intervaltree >=3.0.2          :  3.1.0 (OK)
IPython >=7.6.0               :  7.27.0 (OK)
jedi >=0.17.2;<0.19.0         :  0.18.0 (OK)
jsonschema >=3.2.0            :  3.2.0 (OK)
keyring >=17.0.0              :  22.3.0 (OK)
nbconvert >=4.0               :  5.5.0 (OK)
numpydoc >=0.6.0              :  1.1.0 (OK)
paramiko >=2.4.0              :  2.7.2 (OK)
parso >=0.7.0;<0.9.0          :  0.8.0 (OK)
pexpect >=4.4.0               :  4.8.0 (OK)
pickleshare >=0.4             :  0.7.5 (OK)
psutil >=5.3                  :  5.8.0 (OK)
pygments >=2.0                :  2.7.1 (OK)
pylint >=2.5.0;<2.10.0        :  2.9.6 (OK)
pyls_spyder >=0.4.0           :  0.4.0 (OK)
pylsp >=1.2.2;<1.3.0          :  1.2.3 (OK)
pylsp_black >=1.0.0           :  None (OK)
qdarkstyle =3.0.2             :  3.0.2 (OK)
qstylizer >=0.1.10            :  0.1.10 (OK)
qtawesome >=1.0.2             :  1.0.2 (OK)
qtconsole >=5.1.0             :  5.1.1 (OK)
qtpy >=1.5.0                  :  1.9.0 (OK)
rtree >=0.9.7                 :  0.9.7 (OK)
setuptools >=49.6.0           :  58.2.0 (OK)
sphinx >=0.6.6                :  3.2.1 (OK)
spyder_kernels >=2.1.1;<2.2.0 :  2.1.1 (OK)
textdistance >=4.2.0          :  4.2.1 (OK)
three_merge >=0.1.1           :  0.1.1 (OK)
watchdog >=0.10.3             :  2.1.3 (OK)
zmq >=17                      :  22.2.1 (OK)

# Optional:
cython >=0.21                 :  None (OK)
matplotlib >=2.0.0            :  3.4.3 (OK)
numpy >=1.7                   :  1.20.3 (OK)
pandas >=1.1.1                :  None (OK)
scipy >=0.17.0                :  1.7.1 (OK)
sympy >=0.7.3                 :  None (OK)

[ccordoba12]

it shows the banner with red background

Ok, could you post an image of that to better understand what happens?

[RomanFoell]

Hi dalthviz,

I mainly used conda forge as channel for installation, perhaps this helps, tomorrow I will update this post and tell you which channels and in which order I used them for installation:

.condarc:

default_channels:
  - anaconda
  - r

channels:
  - conda-forge
  - defaults

allow_other_channels: yes
binstar_upload: no
show_channel_urls: yes
notify_outdated_conda: no
report_errors: yes
ssl_verify: true