Docs are little bit outdated for torch logs

[PaliC]

📚 The doc issue

At the end of setup here I assume the output should be some log describing the program rather than saying that the device does not support torch.compile.

https://pytorch.org/tutorials/recipes/torch_logs.html#setup

Suggest a potential alternative/fix

Run the program in the tutorial and update the doc :)

cc @ezyang @chauhang @penguinwu

[rghosh08]

You could propose this fallback behavior in the PyTorch documentation. Include a note in the tutorial that torch.compile might not be supported on all devices, and provide a code snippet like the one above for users with incompatible environments.

Additionally, you might want to include information about how to troubleshoot this issue, such as:

• Checking for CUDA availability
• Verifying that the correct version of PyTorch is installed
• Ensuring proper backend support for torch.compile

Should I work on this?

[PaliC]

@rghosh08 go for it!

[rghosh08]

@PaliC I am trying to raise a PR after signing the CLA: #86170

However, I am getting following error:

` git push --set-upstream origin outdated_torch_logs_137285
ERROR: Permission to pytorch/tutorials.git denied to rghosh08.
fatal: Could not read from remote repository.

Please make sure you have the correct access rights
and the repository exists.`

This is my first contribution to the Pytorch community. Appreciate some hand holding.

[rghosh08]

never mind @PaliC I think I got it. I needed to fork it first.

Here is the PR: pytorch/tutorials#3077

[exekis]

Hey @PaliC! I also made a PR let me know if you want me to make changes :)

[tijikthomas]

@PaliC The reason for this issue seems to be that the tutorial was run during the build process on a CI system that does not support or is not compatible with torch.compile. Is this understanding correct? If so, how can this be fixed in a PR? Would ensuring that the tutorial is tested on a compatible device as part of the CI pipeline resolve the issue? Would a PR be required to trigger the CI and update the output?

(Note: I’m relatively new to contributing to PyTorch, so I appreciate any guidance or clarification.)

[eshentials]

Hi, can I be assigned to this?

[JaynouOliver]

If there is no one working on it, I would like to pick this up please

Hello,
I noticed that previous pull requests addressing this issue (#3077 and #3082) were ultimately closed without being merged. If no one is actively working on a revised solution, I’d like to propose a minimal update.

My intent is to:

• Preserve the CUDA-specific content and structure of the original tutorial
• Add a lightweight fallback or explanatory note for CPU-only environments (as my current setup does not support torch.compile)
• Avoid significant restructuring or behavior changes, in line with previous reviewer feedback

Please let me know if it’s okay to proceed. Happy to coordinate with others if someone else is still planning to take this on.

[dataWizard7957]

Hi I would like to work on this. Please let me know if I can proceed.