VS Jupyter kernel crashes in debug mode

[deep-learning-newbie]

Does this issue occur when all extensions are disabled?: Yes/No

• VS Code Version:
• OS Version:

Steps to Reproduce:

1. Connect VSCode to the remote system using SSH connection
2. enter debug session
3. point mouse at the variable
4. debug session will be crashed and you can see message - Server[1] disconnected unexpectedly
5. the same in jupyter cell debug

My operation system windws 11. remote OS Ubuntu 20.
VSCode version
Version: 1.82.2 (user setup)
Commit: abd2f3db4bdb28f9e95536dfa84d8479f1eb312d
Date: 2023-09-14T05:55:25.390Z
Electron: 25.8.1
ElectronBuildId: 23779380
Chromium: 114.0.5735.289
Node.js: 18.15.0
V8: 11.4.183.29-electron.0
OS: Windows_NT x64 10.0.22621

extensions:

• Jupyter v2023.8.1002501831
• Python v2023.16.0

UPDATE:
the issue appear with tensorflow tensors only. for instance the code below will crash

        gt_boxes = tf.constant([
            [[0, 0, 2, 2], [2, 2, 4, 4]],
            [[0, 0, 2, 2], [2, 2, 4, 4]]
            ], dtype=tf.float32)

        pred_boxes = tf.constant([
            [[0, 0, 2, 2], [2, 2, 4, 4]],
            [[2, 2, 4, 4], [0, 0, 2, 2]]
        ], dtype=tf.float32)

I tested it with different docker images (3 or 4 images) - the same result. currently i use nvcr.io/nvidia/tensorflow:23.08-tf2-py3 image and the issue is reproduces there also .

[roblourens]

I would guess that this is a jupyter/debugpy issue and you would see the same when debugging with jupyterlab? Can you check?

[deep-learning-newbie]

sorry i don't understand what you mean - how can I check it in Jupyter lab- do you mean %whos or locals() %debug or do you mean something else. I can't point mouse at variable and watch its value in Jupyter lab. also I can't expand variables in jyputer - that's why i use vscode. could you provide me instruction what should i do, or install, or setup to be able to provide something useful for you ?

• when I pointed over mouse at variable

• when i expand the variable in watch window will crash

[roblourens]

Would you know anything about this @DonJayamanne?

[deep-learning-newbie]

No, but if you want I install something to test it to figure out the reason of this issue - just let me know. I'm opened because it really hard to develop without debugging. Some time ago it worked ok but now i got this issue and it's make me mad

[DonJayamanne]

@deep-learning-newbie thanks, that wont be necessary, as I too can replicate this at my end. Will look into this.

The only logs we have are

14:13:56.008 [debug] [Debug] to kernel: {"command":"variables","arguments":{"variablesReference":15},"type":"request","seq":41}
14:13:56.008 [debug] [Debug] to kernel, mapped: {"command":"variables","arguments":{"variablesReference":15},"type":"request","seq":41}
14:13:56.098 [debug] KernelProcess Exited, Exit Code - undefined 
14:13:56.098 [error] Disposing session as kernel process died ExitCode: undefined, Reason: 
14:13:56.099 [info] Dispose Kernel process 33599.
14:13:56.102 [debug] Disposed Kernel process 33599.
14:13:56.116 [info] End cell 1 execution @ undefined, started @ 1695269626095, elapsed time = -1695269626.095s
14:13:56.121 [debug] Executing silently Code (dead) = import os as _VSCODE_os\n_VSCODE_fileList = ["/var/folders/fv/524cyb9j03s16612v0f52g4h0000gn/T/ipyker
14:13:56.133 [info] Skipping sending message request because session is disposed

[DonJayamanne]

Crashes in regular Python as well, i.e. this isn't an issue with Jupyter ntoebooks,