Add getter functions for TLM defaults

[AshishSardana]

Key Info

• Implementation plan: link
• Priority:

What changed?

What do you want the reviewer(s) to focus on?

---

Checklist

• Did you link the GitHub issue?
• Did you follow deployment steps or bump the version if needed?
• Did you add/update tests?
• What QA did you do?
  • Tested...

[jwmueller]

needs more tests

[AshishSardana]

A test, unrelated to this PR, is failing.

FAILED tests/test_validation.py::test_custom_eval_criteria_validation - AssertionError: Regex pattern did not match.
 Regex: "^Invalid keys {'extra'} found in custom_eval_criteria item 0. Supported keys are: {'name', 'criteria'}.$"
 Input: "Invalid keys {'extra'} found in custom_eval_criteria item 0. Supported keys are: {'criteria', 'name'}."

It's because of the ordering in {'criteria', 'name'}. Let me know if this is expected (and should let it be), or if I should update this test to work independent of the ordering.

[jwmueller]

A test, unrelated to this PR, is failing.
It's because of the ordering in {'criteria', 'name'}. Let me know if this is expected (and should let it be), or if I should update this test to work independent of the ordering.

You can ignore, unless you have clear idea of how to fix it yourself

[jwmueller]

can you confirm this code will still work:

s = “gpt-4.1-mini”
enc = tiktoken.encoding_for_model(get_default_model(s))

because we're going to shortly be upgrading the default model

[AshishSardana]

The code will not work at the moment because of a pending PR openai/tiktoken#396

Though, its confirmed that 4.1-* uses the same tokenizer as 4o here

When we upgrade to 4.1, we can spoof / hardcode to fetch gpt-4o tokenizer (just setting a variable) until the above PR gets merged.

Let me know if you want me to do this as part of this PR

[jwmueller]

yep I made the suggestion in the code, if you can fix it up / test it works still

[jwmueller]

LGTM assuming these comments are satisfied:

1. https://github.com/cleanlab/cleanlab-tlm/pull/59/files#r2072275920
2. https://github.com/cleanlab/cleanlab-tlm/pull/59/files#r2072275469