File size: 4,745 Bytes
566d03e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
"""
Data Loader
===========

Load benchmark data and evaluation results.
"""

import json
import os
from typing import Dict, List, Optional
from pathlib import Path
from datasets import load_dataset


# Category weights for overall score
CATEGORY_WEIGHTS = {
    "simple": 0.15,
    "multiple": 0.10,
    "parallel": 0.10,
    "parallel_multiple": 0.10,
    "irrelevance": 0.15,
    "dialect_handling": 0.15,
    "multi_turn": 0.15,
    "native_arabic": 0.10,
    # Programming categories (included in evaluation but lower weight)
    "java": 0.0,
    "javascript": 0.0,
    "rest": 0.0,
    "sql": 0.0,
}


def load_benchmark(
    dataset_name: str = "HeshamHaroon/Arabic_Function_Calling",
    split: str = "test",
    category: Optional[str] = None
) -> List[Dict]:
    """
    Load benchmark samples from HuggingFace dataset.

    Args:
        dataset_name: HuggingFace dataset repository
        split: Dataset split ('train' or 'test')
        category: Optional category filter

    Returns:
        List of sample dictionaries
    """
    try:
        dataset = load_dataset(dataset_name, split=split)
    except Exception as e:
        print(f"Error loading dataset: {e}")
        # Fallback to local data
        local_path = Path(__file__).parent.parent.parent / "arabic_fc_dataset" / "checkpoint.json"
        if local_path.exists():
            with open(local_path, 'r', encoding='utf-8') as f:
                data = json.load(f)
            samples = data.get('samples', [])
            if category:
                samples = [s for s in samples if s.get('category') == category]
            return samples
        raise

    samples = []
    for item in dataset:
        sample = {
            'id': item['id'],
            'query_en': item['query_en'],
            'query_ar': item['query_ar'],
            'functions': json.loads(item['functions']) if item['functions'] else [],
            'ground_truth': json.loads(item['ground_truth']) if item['ground_truth'] else None,
            'category': item['category'],
            'source': item.get('source', ''),
            'dialect': item.get('dialect', ''),
        }
        if category is None or sample['category'] == category:
            samples.append(sample)

    return samples


def load_results(results_dir: str = "data/results") -> Dict[str, Dict]:
    """
    Load evaluation results for all models.

    Args:
        results_dir: Directory containing result JSON files

    Returns:
        Dictionary mapping model names to their results
    """
    results = {}
    results_path = Path(results_dir)

    if not results_path.exists():
        return results

    for file_path in results_path.glob("*.json"):
        model_name = file_path.stem
        with open(file_path, 'r', encoding='utf-8') as f:
            results[model_name] = json.load(f)

    return results


def load_leaderboard(leaderboard_path: str = "data/leaderboard.json") -> List[Dict]:
    """
    Load the current leaderboard rankings.

    Returns:
        List of model entries sorted by overall score
    """
    path = Path(leaderboard_path)
    if not path.exists():
        return []

    with open(path, 'r', encoding='utf-8') as f:
        data = json.load(f)

    return sorted(data, key=lambda x: x.get('overall', 0), reverse=True)


def save_leaderboard(entries: List[Dict], leaderboard_path: str = "data/leaderboard.json"):
    """Save leaderboard data to file."""
    path = Path(leaderboard_path)
    path.parent.mkdir(parents=True, exist_ok=True)

    # Sort by overall score
    sorted_entries = sorted(entries, key=lambda x: x.get('overall', 0), reverse=True)

    # Add ranks
    for i, entry in enumerate(sorted_entries, 1):
        entry['rank'] = i

    with open(path, 'w', encoding='utf-8') as f:
        json.dump(sorted_entries, f, ensure_ascii=False, indent=2)


def calculate_overall_score(category_scores: Dict[str, float]) -> float:
    """
    Calculate weighted overall score from category scores.

    Args:
        category_scores: Dictionary mapping category names to scores (0-100)

    Returns:
        Overall weighted score (0-100)
    """
    total_weight = 0
    weighted_sum = 0

    for category, weight in CATEGORY_WEIGHTS.items():
        if category in category_scores and weight > 0:
            weighted_sum += category_scores[category] * weight
            total_weight += weight

    if total_weight == 0:
        return 0.0

    return weighted_sum / total_weight


def get_category_stats(samples: List[Dict]) -> Dict[str, int]:
    """Get sample counts by category."""
    stats = {}
    for sample in samples:
        category = sample.get('category', 'unknown')
        stats[category] = stats.get(category, 0) + 1
    return stats