[ayushdg]

Description

Adds more metrics to track the read, minhash and write time that can provide more complete stats in benchmarks

Closes NMCUR-187

Usage

# Add snippet demonstrating usage

Checklist

• I am familiar with the Contributing Guide.
• New or Existing tests cover these changes.
• The documentation is up to date with these changes.

[greptile-apps]

Greptile Summary

Adds granular custom timing metrics to the MinHash stage.

• Records input read, MinHash computation, and output write durations independently.
• Extends the MinHash stage test to verify all three timing metrics are recorded.

Confidence Score: 5/5

The PR appears safe to merge with no actionable defects identified.

The new timing contexts use the existing stage metric lifecycle, and the backend consumes and aggregates these custom metrics after each task without altering MinHash inputs, outputs, or computation.

Important Files Changed

Filename	Overview
nemo_curator/stages/deduplication/fuzzy/minhash.py	Wraps the existing read, MinHash computation, and parquet write operations in established custom-metric timing contexts without changing processing behavior.
tests/stages/deduplication/fuzzy/test_minhash_stage.py	Verifies that each newly introduced timing metric is populated with a positive duration.

_{Reviews (1): Last reviewed commit: "Add more custom metrics to minhash stage" | Re-trigger Greptile}

[sarahyurick]

LGTM