[TLM] Add TLMCalibrated class

[huiwengoh]

Building off the custom eval criteria in #325

Sample workflow:

prompts = ["what is 1+1", "what is 1+2", "what is 1+3"]
responses = ["2", "hm let me think, i think the answer is 3? am i right?", "5"]
ratings = [5, 3, 1]

custom_eval_criteria_option = {"custom_eval_criteria": [
    {"name": "Conciseness", "criteria": "Determine if the output is concise."}
]}

tlm = studio.TLM(options=custom_eval_criteria_option)
scores = tlm.get_trustworthiness_score(prompts, responses)

tlm_calibrated = studio.TLMCalibrated(options=custom_eval_criteria_option)
tlm_calibrated.fit(scores, ratings)

tlm_calibrated.get_trustworthiness_score("what is 2+2", "5")
>> {'trustworthiness_score': 0.050770278656628066,
 'log': {'custom_eval_criteria': [{'name': 'Conciseness', 'score': 0.8}]},
 'calibrated_score': 0.365}

tlm_calibrated.get_trustworthiness_score(["what is 2+2", "what is 2+3"], ["i think it is 4? maybe 3", "5"])
>> [{'trustworthiness_score': 0.6060140420912912,
  'log': {'custom_eval_criteria': [{'name': 'Conciseness',
     'score': 0.2641642614722914}]},
  'calibrated_score': 0.51},
 {'trustworthiness_score': 0.9967050946426079,
  'log': {'custom_eval_criteria': [{'name': 'Conciseness',
     'score': 0.9999999999999998}]},
  'calibrated_score': 0.79}]

[github-actions]

Ensure final changes made to the TLM code are tested before merging. To run the TLM tests, comment /test-tlm on this PR. To re-run failed property tests, comment /rerun-failed-test-tlm instead.

[github-actions]

Ensure final changes made to the TLM code are tested before merging. To run the TLM tests, comment /test-tlm on this PR. To re-run failed property tests, comment /rerun-failed-test-tlm instead.

[github-actions]

Ensure final changes made to the TLM code are tested before merging. To run the TLM tests, comment /test-tlm on this PR. To re-run failed property tests, comment /rerun-failed-test-tlm instead.

[github-actions]

Ensure final changes made to the TLM code are tested before merging. To run the TLM tests, comment /test-tlm on this PR. To re-run failed property tests, comment /rerun-failed-test-tlm instead.

[github-actions]

Ensure final changes made to the TLM code are tested before merging. To run the TLM tests, comment /test-tlm on this PR. To re-run failed property tests, comment /rerun-failed-test-tlm instead.

[jas2600]

for unit test this new class and the custom eval feature, created ticket https://www.notion.so/cleanlab/Add-unit-tests-in-cleanlab-studio-for-TLMCalibrated-and-the-custom-eval-feature-c01d32f159c7473a878a998b37f52764

[github-actions]

Ensure final changes made to the TLM code are tested before merging. To run the TLM tests, comment /test-tlm on this PR. To re-run failed property tests, comment /rerun-failed-test-tlm instead.

[jas2600]

/test-tlm
✨ Starting TLM tests... ✨
If you want to run all the TLM tests again (because TLM code is ready for review), comment '/test-tlm' on this PR.
If you want to re-run only the failed tests (you are still developing), comment '/rerun-failed-test-tlm' on this PR.
View full GitHub Actions run log
✨ Tests completed! ✨
TLM Tests Results: ✅✅✅✅✅
TLM Property Tests Results: ✅✅✅✅✅
Click the Github Actions run log for more information.

[github-actions]

Ensure final changes made to the TLM code are tested before merging. To run the TLM tests, comment /test-tlm on this PR. To re-run failed property tests, comment /rerun-failed-test-tlm instead.