fix: raise ValueError when provided seed is out-of-bounds

[jonathanking]

What does this PR do?

Changes behavior of seed_everything to raise a ValueError if and only if the user provides a seed that is out of bounds ([0, 2**32-1]) in order to more clearly match the user's expectations. Prevents misconfiguration of seeds when highly predictable behavior is desired.

This behavior is changed whether the user specified the seed via the seed_everything function call or the PL_GLOBAL_SEED env var.

Fixes #21028
See also #18846 for context.
cc @deependujha

Breaking changes:

1. If a user provides an integer to seed_everything that is outside of [0, 2**32-1], the code will raise a ValueError where it previously changed the seed to 0.

Before submitting

• Was this discussed/agreed via a GitHub issue? (not for typos and docs)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure your PR does only one thing, instead of bundling different changes together?
• Did you make sure to update the documentation with your changes? (if necessary)
• Did you write any new necessary tests? (not for typos and docs)
• Did you verify new and existing tests pass locally with your changes?
• Did you list all the breaking changes introduced by this pull request?
• Did you update the CHANGELOG? (not for typos, docs, test updates, or minor internal changes/refactors)

PR review

Anyone in the community is welcome to review the PR.
Before you start reviewing, make sure you have read the review guidelines. In short, see the following bullet-list:

Reviewer checklist

• Is this pull request ready for review? (if not, please submit in draft mode)
• Check that all items from Before submitting are resolved
• Make sure the title is self-explanatory and the description concisely explains the PR
• Add labels and milestones (and optionally projects) to the PR so it can be classified

---

📚 Documentation preview 📚: https://pytorch-lightning--21029.org.readthedocs.build/en/21029/

[deependujha]

do you intend to make more changes, or should we mark it as ready for review?

[jonathanking]

Thanks for reviewing! I have one question to resolve:

I have only changed the behavior to raise a ValueError if a user calls seed_everything with an invalid integer seed. However, I did not change the behavior for if the user specifies an invalid seed via the environment variable.

The benefit of my current implementation is that it changes as little code as possible and mostly maintains the original behavior. The downside of my current implementation is that the behavior is different depending on whether or not the user uses the environment variable or the function argument.

Having thought about this a bit more, I propose to change the behavior also for the case that the user specifies the seed via the environment variable. What do you think?

[deependujha]

Hi @jonathanking

sorry, I only looked at diff. Let's be consistent. Raise ValueError in env var case too.

[deependujha]

can you include test for it as well (with env variable)

[jonathanking]

Thanks, I agree - just updated the code and test.

[jonathanking]

Thanks for the feedback, @deependujha! Will now mark as ready.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 87%. Comparing base (e752cb5) to head (90d88f9).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master   #21029   +/-   ##
=======================================
- Coverage      87%      87%   -0%     
=======================================
  Files         268      268           
  Lines       23303    23301    -2     
=======================================
- Hits        20297    20295    -2     
  Misses       3006     3006           

[Copilot]

Pull Request Overview

This PR changes the behavior of seed_everything to raise a ValueError when an invalid or out-of-bounds seed is provided, instead of falling back to a default seed value of 0 with a warning. This makes the function more strict and prevents misconfiguration when predictable seeding behavior is critical.

• Replace warning-based fallback behavior with strict error handling for invalid seeds
• Update error messages to be more descriptive and consistent
• Add comprehensive test coverage for the new validation behavior

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
src/lightning/fabric/utilities/seed.py	Updated seed validation logic to raise ValueError instead of warnings with fallback to 0
tests/tests_fabric/utilities/test_seed.py	Updated existing tests and added new tests to verify ValueError is raised for invalid seeds
src/lightning/fabric/CHANGELOG.md	Added changelog entry documenting the breaking change

[bhimrazy]

Thanks @jonathanking for the update! I agree with the intent here.

Curious to hear what @Borda, @SkafteNicki, and @matsumotosan have to say on this

[deependujha]

Thanks @jonathanking for helping make pytorch lightning even better.