Skip to content

test: cover HumanEval evaluator details #2622

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
motodriver wants to merge 1 commit into
base: main
Choose a base branch
from
+172 −1
Open

test: cover HumanEval evaluator details #2622

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
173 changes: 172 additions & 1 deletion tests/datasets/test_humaneval.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,13 @@
import json
import os
import sys
import unittest
from types import ModuleType
from unittest.mock import Mock, patch

from opencompass.datasets.humaneval import (humaneval_postprocess_v2,
from opencompass.datasets.humaneval import (HumanEvalEvaluator,
HumanEvalPlusEvaluator,
humaneval_postprocess_v2,
humaneval_postprocess_v3)


Expand Down Expand Up @@ -73,5 +80,169 @@ def test_v3_matches_v2_for_single_code_block(self):
self.assertEqual(humaneval_postprocess_v3(raw), expected)


class TestHumanEvalEvaluatorDetails(unittest.TestCase):

def test_score_returns_details_from_mocked_evaluation_results(self):
human_eval = ModuleType('human_eval')
human_eval_data = ModuleType('human_eval.data')
human_eval_evaluation = ModuleType('human_eval.evaluation')
write_jsonl = Mock()

def fake_evaluate(sample_file, *args, **kwargs):
result_file = sample_file + '_results.jsonl'
rows = [
{
'task_id': 'HumanEval/0',
'completion': 'return 1',
'passed': True,
'result': 'passed',
},
{
'task_id': 'HumanEval/1',
'completion': 'return 0',
'passed': False,
'result': 'failed: assertion',
},
]
with open(result_file, 'w', encoding='utf-8') as result:
for row in rows:
result.write(json.dumps(row) + '\n')
return {'pass@1': 0.5}

evaluate = Mock(side_effect=fake_evaluate)
human_eval_data.HUMAN_EVAL = 'mock_problem_file.jsonl'
human_eval_data.write_jsonl = write_jsonl
human_eval_evaluation.evaluate_functional_correctness = evaluate

modules = {
'human_eval': human_eval,
'human_eval.data': human_eval_data,
'human_eval.evaluation': human_eval_evaluation,
}
with patch.dict(sys.modules, modules):
result = HumanEvalEvaluator(k=[1]).score(
predictions=['return 1', 'return 0'],
references=['HumanEval/0', 'HumanEval/1'],
test_set=[{
'prompt': 'def one():'
}, {
'prompt': 'def two():'
}],
)

self.assertEqual(result['humaneval_pass@1'], 50.0)
self.assertEqual(
result['details'], {
'0': {
'task_id': 'HumanEval/0',
'completion': 'return 1',
'passed': True,
'result': 'passed',
'is_correct': True,
'prompt': 'def one():',
},
'1': {
'task_id': 'HumanEval/1',
'completion': 'return 0',
'passed': False,
'result': 'failed: assertion',
'is_correct': False,
'prompt': 'def two():',
},
})
sample_file, samples = write_jsonl.call_args.args
self.assertEqual(samples, [{
'task_id': 'HumanEval/0',
'completion': 'return 1'
}, {
'task_id': 'HumanEval/1',
'completion': 'return 0'
}])
evaluate.assert_called_once_with(sample_file, [1],
n_workers=4,
timeout=3.0,
problem_file='mock_problem_file.jsonl')
self.assertFalse(os.path.exists(os.path.dirname(sample_file)))


class TestHumanEvalPlusEvaluatorDetails(unittest.TestCase):

def test_score_returns_details_from_mocked_evaluation_results(self):
evalplus = ModuleType('evalplus')
evalplus_data = ModuleType('evalplus.data')
evalplus_evaluate = ModuleType('evalplus.evaluate')
write_jsonl = Mock()

def fake_evaluate(flags):
result_file = '{}_eval_results.json'.format(flags.samples[:-6])
eval_results = {
'eval': {
'HumanEval/0': [{
'base_status': 'pass',
'plus_status': 'pass',
}],
'HumanEval/1': [{
'base_status': 'pass',
'plus_status': 'fail',
}],
}
}
with open(result_file, 'w', encoding='utf-8') as result:
json.dump(eval_results, result)

evaluate = Mock(side_effect=fake_evaluate)
evalplus_data.write_jsonl = write_jsonl
evalplus_evaluate.evaluate = evaluate

modules = {
'evalplus': evalplus,
'evalplus.data': evalplus_data,
'evalplus.evaluate': evalplus_evaluate,
}
with patch.dict(sys.modules, modules):
result = HumanEvalPlusEvaluator(k=[1]).score(
predictions=[' return 1', ' return 0'],
references=['HumanEval/0', 'HumanEval/1'],
test_set=[{
'prompt': 'def one():\n'
}, {
'prompt': 'def two():\n'
}],
)

self.assertEqual(result['humaneval_plus_pass@1'], 50.0)
self.assertEqual(
result['details'], {
'0': {
'prompt': 'def one():\n',
'prediction': ' return 1',
'reference': 'HumanEval/0',
'base_result': 'pass',
'plus_result': 'pass',
'is_correct': True,
},
'1': {
'prompt': 'def two():\n',
'prediction': ' return 0',
'reference': 'HumanEval/1',
'base_result': 'pass',
'plus_result': 'fail',
'is_correct': False,
},
})
sample_file, samples = write_jsonl.call_args.args
self.assertEqual(samples, [{
'task_id': 'HumanEval/0',
'solution': 'def one():\n return 1'
}, {
'task_id': 'HumanEval/1',
'solution': 'def two():\n return 0'
}])
flags = evaluate.call_args.args[0]
self.assertEqual(flags.dataset, 'humaneval')
self.assertEqual(flags.samples, sample_file)
self.assertFalse(os.path.exists(os.path.dirname(sample_file)))


if __name__ == '__main__':
unittest.main()