Move dsp/metrics into evaluate/metrics

[TomeHirata]

We had dsp/utils/metrics and evaluate/metrics separately and re-export some metrics under dsp/utils/metrics in evaluate/metrics. To have a clear separation of concerns, this PR moves functions in dsp/utils/metrics into evaluate/metrics.

[Copilot]

Pull Request Overview

Centralize metric implementations by moving functions from dspy/dsp/utils/metrics.py into dspy/evaluate/metrics.py and updating exports.

• Copied EM, F1, HotPotF1, normalize_text, em_score, f1_score, hotpot_f1_score, and precision_score into evaluate/metrics.py
• Updated dspy/evaluate/__init__.py to re-export relocated metrics and removed old imports
• Deleted the old metrics module and cleaned up imports in dspy/dsp/utils/__init__.py

Reviewed Changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

File	Description
dspy/evaluate/metrics.py	Added and integrated core metric functions into the evaluation package
dspy/evaluate/init.py	Updated top-level exports for relocated metrics functions
dspy/dsp/utils/metrics.py	Removed legacy metrics module
dspy/dsp/utils/init.py	Removed metrics import from utils initialization

Comments suppressed due to low confidence (3)

dspy/evaluate/metrics.py:29

• Public function normalize_text lacks a docstring. Add a brief description of its behavior and parameters.

def normalize_text(s):

dspy/evaluate/init.py:3

• Functions F1 and HotPotF1 are defined in metrics.py but not re-exported here. If they’re part of the public API, add them to the imports and __all__.

from dspy.evaluate.metrics import EM, answer_exact_match, answer_passage_match, normalize_text

dspy/evaluate/metrics.py:99

• New function precision_score has no associated unit tests. Consider adding tests for normal and edge cases (e.g., empty inputs) to validate behavior.

def precision_score(prediction, ground_truth):

[TomeHirata]

The change here is for avoiding circular imports