The kernel appears to have died. It will restart automatically. v2

Hi,

I get the message “The kernel appears to have died. It will restart automatically.” every time when I run cells containing learn.fine_tune( ) or learn.fit_one_cycle() in jupyter notebook (or vs code) on my laptop with the following warnings:

my setup is:
|Windows 10 Pro|Version 22H2|OS build 19045.2251|
WSL2 Ubuntu 22.04.1 LTS

=== Software === 
python        : 3.10.8
fastai        : 2.7.10
fastcore      : 1.5.27
fastprogress  : 1.0.3
torch         : 1.13.0
nvidia driver : 517.00
torch cuda    : 11.7 / is available
torch cudnn   : 8500 / is enabled

=== Hardware === 
nvidia gpus   : 1
torch devices : 1
  - gpu0      : Quadro T2000

=== Environment === 
platform      : Linux-5.10.102.1-microsoft-standard-WSL2-x86_64-with-glibc2.35
distro        : #1 SMP Wed Mar 2 00:30:59 UTC 2022
conda env     : base
python        : /home/ts/mambaforge/envs/torch/bin/python
sys.path      : /mnt/c/dropbox/ds/fast_ai/fastai_course22
/home/ts/mambaforge/envs/torch/lib/python310.zip
/home/ts/mambaforge/envs/torch/lib/python3.10
/home/ts/mambaforge/envs/torch/lib/python3.10/lib-dynload

/home/ts/.local/lib/python3.10/site-packages
/home/ts/mambaforge/envs/torch/lib/python3.10/site-packages
/home/ts/mambaforge/envs/torch/lib/python3.10/site-packages/sentencepiece-0.1.96-py3.10-linux-x86_64.egg

What I’ve tried so far:

  • create from scratch new image Ubuntu using the instructions from LIve Coding videos
  • downgrade Python to 3.9
  • downgrade PyTorch to 1.12
  • downgrade CUDA to 11.6
  • reinstall/update Nvidia drivers

On the desktop, the same image of Ubuntu works well with the same warnings.

Could someone tell me the reason?
Thank you.

Downgrading to PyTorch 1.11.0 & torchvision 0.12.0 has solved the issue.

2 Likes