fix(triviaqa): update dataset loading path to use correct repository

[cris96spa]

Summary

Fix TriviaQA dataset loading by switching from the legacy dataset identifier to the current namespaced Hugging Face repository ID.

Problem

The loader currently uses load_dataset("trivia_qa", ...). Hugging Face Hub now enforces namespace/name repository IDs, so this fails with repository ID validation errors.

Change

• Update dataset ID from trivia_qa to mandarjoshi/trivia_qa.

Impact

• Restores TriviaQA data loading without changing evaluator semantics.
• Keeps existing seeding and verification behavior intact.

Validation

• Verified dataset streaming and row access using mandarjoshi/trivia_qa.
• Verified expected fields remain compatible with current loader logic.
• Run standard checks:
  • make test
  • make lint

Summary by CodeRabbit

• Bug Fixes
  • Updated TriviaQA benchmark loading to use a more reliable dataset source, helping ensure the benchmark loads correctly.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[coderabbitai]

📝 Walkthrough

Walkthrough

The Hugging Face dataset identifier used in TriviaQA loading is changed from trivia_qa to mandarjoshi/trivia_qa, retaining the same configuration and split. No other logic or public interfaces are altered.

Changes

TriviaQA loader update

Layer / File(s)	Summary
Update dataset source identifier src/nemo_evaluator/benchmarks/triviaqa.py	_load_triviaqa() now loads from mandarjoshi/trivia_qa instead of trivia_qa, keeping the rc.nocontext config and validation split unchanged.

Estimated code review effort: 1 (Trivial) | ~2 minutes

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly describes the main change: updating TriviaQA dataset loading to the correct repository path.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@src/nemo_evaluator/benchmarks/triviaqa.py`:
- Line 33: Pin the TriviaQA dataset load in triviaqa.py for reproducibility by
adding a fixed revision to the load_dataset call in the benchmark setup. Update
the ds assignment in the TriviaQA evaluator so the dataset source is tied to a
specific revision hash rather than the floating default, keeping future runs
stable even if mandarjoshi/trivia_qa changes upstream.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Enterprise

Run ID: ac7b846e-da00-4fbe-a793-816641beaad2

📥 Commits

Reviewing files that changed from the base of the PR and between a668af9 and 88e99a7.

📒 Files selected for processing (1)

• src/nemo_evaluator/benchmarks/triviaqa.py

[coderabbitai]

🗄️ Data Integrity & Integration | 🔵 Trivial | ⚡ Quick win

Consider pinning a dataset revision for reproducibility.

Without a pinned revision, future upstream changes to mandarjoshi/trivia_qa (re-uploads, config renames) could silently change eval results between runs.

♻️ Pin a revision hash

-    ds = load_dataset("mandarjoshi/trivia_qa", "rc.nocontext", split="validation")
+    ds = load_dataset("mandarjoshi/trivia_qa", "rc.nocontext", split="validation", revision="<commit_hash>")

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/nemo_evaluator/benchmarks/triviaqa.py` at line 33, Pin the TriviaQA
dataset load in triviaqa.py for reproducibility by adding a fixed revision to
the load_dataset call in the benchmark setup. Update the ds assignment in the
TriviaQA evaluator so the dataset source is tied to a specific revision hash
rather than the floating default, keeping future runs stable even if
mandarjoshi/trivia_qa changes upstream.

[cris96spa]

Closes #1111