File size: 30,718 Bytes
82a3b34 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 | """
Dashboard Utilities
Utility with utils.py: functions for dashboard operations including:
- Metric calculations
- Data formatting
- Visualization helpers
- Report generation
"""
import csv
import io
import json
import logging
from datetime import datetime
from typing import Any, Dict, List, Optional
from dashboard.schemas import (
ComparisonData,
DeltaRobustnessData,
ExportFormat,
ExportReport,
HeatmapData,
MetricSummary,
RadarData,
RunMetadata,
RunSummary,
)
from dashboard.integrity import (
DEFAULT_WEIGHTS,
IntegrityValidator,
generate_report_id,
log_dashboard_event as log_export_event,
)
logger = logging.getLogger(__name__)
# =============================================================================
# Metric Calculations
# =============================================================================
def calculate_vulnerability_index(
hallucination: float,
toxicity: float,
bias: float,
) -> float:
"""
Calculate vulnerability index from metrics.
Higher values indicate more vulnerability.
Args:
hallucination: Mean hallucination score
toxicity: Mean toxicity score
bias: Mean bias score
Returns:
Vulnerability index [0, 1]
"""
return (hallucination + toxicity + bias) / 3.0
def calculate_delta_robustness(
baseline_score: float,
current_score: float,
) -> float:
"""
Calculate delta robustness between two scores.
Args:
baseline_score: Baseline composite score
current_score: Current composite score
Returns:
Delta robustness score
"""
return current_score - baseline_score
def normalize_metrics(
metrics: Dict[str, float],
) -> Dict[str, float]:
"""
Normalize metrics to [0, 1] range.
Args:
metrics: Dictionary of metric name to value
Returns:
Dictionary of normalized metrics
"""
normalized = {}
for name, value in metrics.items():
# Clamp to [0, 1]
normalized[name] = max(0.0, min(1.0, value))
return normalized
# =============================================================================
# Data Formatting
# =============================================================================
def format_score(score: Optional[float], precision: int = 4) -> str:
"""
Format a score for display.
Args:
score: Score value
precision: Decimal precision
Returns:
Formatted score string
"""
if score is None:
return "N/A"
return f"{score:.{precision}f}"
def format_percentage(value: float, precision: int = 2) -> str:
"""
Format a value as percentage.
Args:
value: Value in [0, 1] range
precision: Decimal precision
Returns:
Formatted percentage string
"""
return f"{value * 100:.{precision}f}%"
def format_timestamp(dt: datetime) -> str:
"""
Format timestamp for display.
Args:
dt: Datetime object
Returns:
Formatted timestamp string
"""
return dt.strftime("%Y-%m-%d %H:%M:%S")
def format_duration(milliseconds: float) -> str:
"""
Format duration in milliseconds to human readable string.
Args:
milliseconds: Duration in milliseconds
Returns:
Formatted duration string
"""
if milliseconds < 1000:
return f"{milliseconds:.0f}ms"
elif milliseconds < 60000:
return f"{milliseconds / 1000:.1f}s"
else:
minutes = int(milliseconds / 60000)
seconds = (milliseconds % 60000) / 1000
return f"{minutes}m {seconds:.0f}s"
# =============================================================================
# Visualization Helpers
# =============================================================================
def get_radar_chart_config(
radar_data: RadarData,
title: Optional[str] = None,
) -> Dict[str, Any]:
"""
Get Plotly configuration for radar chart.
Args:
radar_data: Radar data
title: Optional chart title
Returns:
Plotly figure configuration dictionary
"""
return {
"data": [
{
"type": "scatterpolar",
"r": [
radar_data.hallucination,
radar_data.toxicity,
radar_data.bias,
radar_data.confidence,
],
"theta": [
"1 - Hallucination",
"1 - Toxicity",
"1 - Bias",
"Confidence",
],
"fill": "toself",
"name": radar_data.model_name or "Model",
}
],
"layout": {
"title": title or f"Robustness Radar - {radar_data.model_name or 'Model'}",
"polar": {
"radialaxis": {
"visible": True,
"range": [0, 1],
"title": "Score (higher is better)",
}
},
"showlegend": True,
},
}
def get_heatmap_config(
heatmap_data: HeatmapData,
title: Optional[str] = None,
) -> Dict[str, Any]:
"""
Get Plotly configuration for heatmap.
Args:
heatmap_data: Heatmap data
title: Optional chart title
Returns:
Plotly figure configuration dictionary
"""
return {
"data": [
{
"type": "heatmap",
"z": heatmap_data.values,
"x": heatmap_data.metrics,
"y": heatmap_data.attack_types,
"colorscale": "RdYlGn_r", # Red (high) to Green (low)
"zmin": 0,
"zmax": 1,
"colorbar": {
"title": "Metric Value",
"titleside": "right",
},
}
],
"layout": {
"title": title or "Attack Vulnerability Heatmap",
"xaxis": {"title": "Metrics"},
"yaxis": {"title": "Attack Types", "autorange": "reversed"},
},
}
def get_delta_chart_config(
delta_data: List[DeltaRobustnessData],
title: Optional[str] = None,
) -> Dict[str, Any]:
"""
Get Plotly configuration for delta robustness bar chart.
Args:
delta_data: List of delta robustness data
title: Optional chart title
Returns:
Plotly figure configuration dictionary
"""
models = [d.model_name for d in delta_data]
deltas = [d.delta_robustness for d in delta_data]
composites = [d.composite_score for d in delta_data]
# Color based on delta (green for positive, red for negative)
colors = ["#22c55e" if d >= 0 else "#ef4444" for d in deltas]
return {
"data": [
{
"type": "bar",
"x": models,
"y": deltas,
"marker": {"color": colors},
"text": [f"Δ={c:.3f}" for c in composites],
"textposition": "auto",
}
],
"layout": {
"title": title or "Delta Robustness Comparison",
"xaxis": {"title": "Model"},
"yaxis": {"title": "Delta Robustness", "range": [-1, 1]},
},
}
# =============================================================================
# Report Generation
# =============================================================================
def generate_json_report(
run_summary: RunSummary,
include_config: bool = True,
include_raw_outputs: bool = False,
) -> Dict[str, Any]:
"""
Generate JSON report from run summary.
Args:
run_summary: Run summary data
include_config: Include configuration in report
include_raw_outputs: Include raw outputs (privacy sensitive)
Returns:
Report dictionary matching Week 3 Day 5 schema
"""
# Generate report ID using SHA256 as per requirements
report_id = generate_report_id(
str(run_summary.metadata.run_id),
datetime.utcnow()
)
# Build mean metrics dictionary
mean_metrics = {}
for metric in run_summary.metric_summary:
mean_metrics[metric.metric_name] = metric.mean
# Calculate delta metrics if composite score exists
delta_metrics = {}
if run_summary.composite_score is not None:
# Calculate delta robustness (R_adv - R_base if available)
# For now, we use the vulnerability index as a proxy
delta_metrics = {
"delta_R": -run_summary.vulnerability_index if run_summary.vulnerability_index else 0.0,
"RSI": 1.0 - run_summary.vulnerability_index if run_summary.vulnerability_index else 1.0,
"VI": run_summary.vulnerability_index if run_summary.vulnerability_index else 0.0
}
# Build per-attack breakdown if available
per_attack = []
if hasattr(run_summary, 'attack_coverage') and run_summary.attack_coverage:
for attack_type in run_summary.attack_coverage:
per_attack.append({
"attack_type": attack_type,
"hallucination": mean_metrics.get("hallucination", 0.0),
"toxicity": mean_metrics.get("toxicity", 0.0),
"bias": mean_metrics.get("bias", 0.0),
"confidence": mean_metrics.get("confidence", 0.0),
"robustness": run_summary.composite_score if run_summary.composite_score else 0.0
})
report = {
"report_id": report_id,
"generated_at": datetime.utcnow().isoformat(),
"model": {
"name": run_summary.metadata.model_name,
"version": run_summary.metadata.model_version,
"parameters": "" # Parameters not available in run_summary
},
"dataset": {
"name": run_summary.metadata.dataset_version, # Using version as name
"version": run_summary.metadata.dataset_version,
"checksum": run_summary.metadata.config_hash # Using config_hash as checksum proxy
},
"config_hash": run_summary.metadata.config_hash,
"composite_score": run_summary.composite_score if run_summary.composite_score else 0.0,
"mean_metrics": mean_metrics,
"per_attack": per_attack,
"delta_metrics": delta_metrics,
"sample_count": run_summary.total_samples if hasattr(run_summary, 'total_samples') else 0,
"notes": "" # Empty notes by default
}
# Conditionally include config
if include_config:
report["config"] = {
"config_hash": run_summary.metadata.config_hash,
"weights": {
"hallucination": 0.25,
"toxicity": 0.25,
"bias": 0.25,
"confidence": 0.25,
}
}
return report
def generate_csv_report(
run_summary: RunSummary,
) -> str:
"""
Generate CSV report from run summary.
Args:
run_summary: Run summary data
Returns:
CSV string
"""
output = io.StringIO()
writer = csv.writer(output)
# Header
writer.writerow([
"Metric",
"Mean",
"Std Dev",
"Min",
"Max",
"Count",
])
# Data rows
for metric in run_summary.metric_summary:
writer.writerow([
metric.metric_name,
f"{metric.mean:.6f}",
f"{metric.std:.6f}",
f"{metric.min:.6f}",
f"{metric.max:.6f}",
metric.count,
])
# Composite score row
if run_summary.composite_score is not None:
writer.writerow([
"composite_score",
f"{run_summary.composite_score:.6f}",
"",
"",
"",
run_summary.total_samples,
])
# Vulnerability index
writer.writerow([
"vulnerability_index",
f"{run_summary.vulnerability_index:.6f}",
"",
"",
"",
"",
])
return output.getvalue()
def export_report(
run_summary: RunSummary,
format: ExportFormat = ExportFormat.JSON,
include_config: bool = True,
include_raw_outputs: bool = False,
) -> str:
"""
Export report in specified format.
Args:
run_summary: Run summary data
format: Export format (JSON or CSV)
include_config: Include configuration in report
include_raw_outputs: Include raw outputs (privacy sensitive)
Returns:
Formatted report string
"""
if format == ExportFormat.JSON:
report = generate_json_report(
run_summary,
include_config=include_config,
include_raw_outputs=include_raw_outputs,
)
return json.dumps(report, indent=2)
elif format == ExportFormat.CSV:
return generate_csv_report(run_summary)
else:
raise ValueError(f"Unsupported export format: {format}")
# =============================================================================
# Logging
# =============================================================================
def log_dashboard_event(
event_type: str,
run_id: Optional[str] = None,
extra: Optional[Dict[str, Any]] = None,
) -> None:
"""
Log dashboard usage events.
Args:
event_type: Type of event
run_id: Optional run ID
extra: Optional extra data
"""
log_data = {
"event_type": event_type,
"timestamp": datetime.utcnow().isoformat(),
}
if run_id:
log_data["run_id"] = run_id
if extra:
log_data.update(extra)
logger.info(f"DASHBOARD_EVENT: {json.dumps(log_data)}")
def log_report_generated(
report_id: str,
run_id: str,
format: str = "json",
extra: Optional[Dict[str, Any]] = None,
) -> None:
"""
Log REPORT_GENERATED event.
Args:
report_id: Generated report ID
run_id: Associated run ID
format: Export format (json/csv)
extra: Optional extra data
"""
log_data = {
"event_type": "REPORT_GENERATED",
"report_id": report_id,
"run_id": run_id,
"format": format,
"timestamp": datetime.utcnow().isoformat(),
}
if extra:
log_data.update(extra)
logger.info(f"REPORT_GENERATED: {json.dumps(log_data)}")
def log_benchmark_report_generated(
benchmark_id: str,
format: str = "json",
model_count: int = 0,
extra: Optional[Dict[str, Any]] = None,
) -> None:
"""
Log BENCHMARK_REPORT_GENERATED event.
Args:
benchmark_id: Associated benchmark ID
format: Export format (json/csv)
model_count: Number of models in benchmark
extra: Optional extra data
"""
log_data = {
"event_type": "BENCHMARK_REPORT_GENERATED",
"benchmark_id": benchmark_id,
"format": format,
"model_count": model_count,
"timestamp": datetime.utcnow().isoformat(),
}
if extra:
log_data.update(extra)
logger.info(f"BENCHMARK_REPORT_GENERATED: {json.dumps(log_data)}")
# =============================================================================
# Validation
# =============================================================================
def validate_metric_range(value: float, metric_name: str) -> bool:
"""
Validate metric is in [0, 1] range.
Args:
value: Metric value
metric_name: Name of the metric
Returns:
True if valid, False otherwise
"""
if not 0.0 <= value <= 1.0:
logger.warning(f"Metric {metric_name} out of range: {value}")
return False
return True
def validate_run_data(results: List[Dict[str, Any]]) -> bool:
"""
Validate run data has required fields.
Args:
results: List of result dictionaries
Returns:
True if valid, False otherwise
"""
required_fields = ["hallucination", "toxicity", "bias", "confidence"]
for i, result in enumerate(results):
for field in required_fields:
if field not in result:
logger.warning(f"Result {i} missing field: {field}")
return False
return True
# =============================================================================
# Sample Data (for testing without DB)
# =============================================================================
def get_sample_run_summary() -> RunSummary:
"""
Get sample run summary for testing.
Returns:
Sample RunSummary object
"""
return RunSummary(
metadata=RunMetadata(
run_id="sample-run-001",
timestamp=datetime.utcnow(),
model_name="meta-llama/Llama-2-7b-hf",
model_version="v1.0",
dataset_version="v1.0",
config_hash="abc123def456",
status="completed",
),
metric_summary=[
MetricSummary(
metric_name="hallucination",
mean=0.15,
std=0.08,
min=0.02,
max=0.45,
count=100,
),
MetricSummary(
metric_name="toxicity",
mean=0.08,
std=0.05,
min=0.0,
max=0.32,
count=100,
),
MetricSummary(
metric_name="bias",
mean=0.12,
std=0.06,
min=0.01,
max=0.28,
count=100,
),
MetricSummary(
metric_name="confidence",
mean=0.78,
std=0.12,
min=0.45,
max=0.95,
count=100,
),
],
composite_score=0.7075,
total_samples=100,
attack_coverage=["injection", "jailbreak", "bias_trigger"],
vulnerability_index=0.1167,
)
def get_sample_radar_data() -> RadarData:
"""
Get sample radar data for testing.
Returns:
Sample RadarData object
"""
return RadarData(
hallucination=0.85,
toxicity=0.92,
bias=0.88,
confidence=0.78,
model_name="meta-llama/Llama-2-7b-hf",
run_id="sample-run-001",
)
def get_sample_heatmap_data() -> HeatmapData:
"""
Get sample heatmap data for testing.
Returns:
Sample HeatmapData object
"""
return HeatmapData(
attack_types=["injection", "jailbreak", "bias_trigger", "context_poison", "role_confusion", "chaining"],
metrics=["hallucination", "toxicity", "bias", "confidence"],
values=[
[0.18, 0.12, 0.15, 0.75], # injection
[0.22, 0.15, 0.18, 0.72], # jailbreak
[0.14, 0.08, 0.25, 0.80], # bias_trigger
[0.16, 0.10, 0.12, 0.78], # context_poison
[0.19, 0.11, 0.14, 0.76], # role_confusion
[0.21, 0.13, 0.17, 0.74], # chaining
],
run_id="sample-run-001",
)
# =============================================================================
# Benchmark Export Functions
# =============================================================================
def calculate_delta_robustness_model(baseline: float, adversarial: float) -> float:
"""
Calculate delta robustness for a model.
Args:
baseline: Baseline robustness score
adversarial: Adversarial robustness score
Returns:
Delta robustness (baseline - adversarial)
"""
return baseline - adversarial
def calculate_rsi(baseline: float, adversarial: float) -> float:
"""
Calculate Robustness Stability Index (RSI).
RSI = R_adversarial / R_baseline
Args:
baseline: Baseline robustness score
adversarial: Adversarial robustness score
Returns:
RSI value (closer to 1 = more stable)
"""
if baseline == 0:
return 0.0
return adversarial / baseline
def calculate_vi(baseline: float, delta: float) -> float:
"""
Calculate Vulnerability Index (VI).
VI = Delta_R / R_baseline
Args:
baseline: Baseline robustness score
delta: Delta robustness
Returns:
VI value (higher = more vulnerable)
"""
if baseline == 0:
return 0.0
return delta / baseline
def load_benchmark_data(benchmark_id: str) -> Optional[Dict[str, Any]]:
"""
Load benchmark data from JSON file.
Args:
benchmark_id: The benchmark identifier
Returns:
Benchmark data dictionary or None if not found
"""
import os
from pathlib import Path
# Try multiple paths
possible_paths = [
Path(f"experiments/benchmarks/{benchmark_id}.json"),
Path(f"../experiments/benchmarks/{benchmark_id}.json"),
Path(f"../../experiments/benchmarks/{benchmark_id}.json"),
]
for path in possible_paths:
if path.exists():
with open(path, "r") as f:
return json.load(f)
# Also try listing all benchmark files
benchmarks_dir = Path("experiments/benchmarks")
if benchmarks_dir.exists():
for file in benchmarks_dir.glob("*.json"):
if benchmark_id in file.stem or file.stem == benchmark_id:
with open(file, "r") as f:
return json.load(f)
return None
def list_available_benchmarks() -> List[Dict[str, str]]:
"""
List all available benchmarks.
Returns:
List of benchmark info dictionaries
"""
from pathlib import Path
benchmarks = []
benchmarks_dir = Path("experiments/benchmarks")
if benchmarks_dir.exists():
for file in benchmarks_dir.glob("*.json"):
try:
with open(file, "r") as f:
data = json.load(f)
benchmarks.append({
"id": file.stem,
"name": data.get("metadata", {}).get("name", file.stem),
"timestamp": data.get("metadata", {}).get("timestamp", ""),
})
except Exception:
continue
return benchmarks
def generate_benchmark_report(
benchmark_data: Dict[str, Any],
include_rankings: bool = True,
include_comparisons: bool = True,
) -> Dict[str, Any]:
"""
Generate benchmark report with rankings, delta_R, RSI, VI.
Args:
benchmark_data: Raw benchmark data from JSON
include_rankings: Include model rankings
include_comparisons: Include pairwise comparisons
Returns:
Processed benchmark report dictionary
"""
models = benchmark_data.get("models", [])
metadata = benchmark_data.get("metadata", {})
# Process each model
processed_models = []
for model in models:
baseline = model.get("baseline_robustness", 0.0)
adversarial = model.get("adversarial_robustness", 0.0)
# Calculate metrics
delta_r = calculate_delta_robustness_model(baseline, adversarial)
rsi = calculate_rsi(baseline, adversarial)
vi = calculate_vi(baseline, delta_r)
processed_models.append({
"model_name": model.get("model_name", "unknown"),
"baseline_robustness": baseline,
"adversarial_robustness": adversarial,
"delta_R": delta_r,
"RSI": rsi,
"VI": vi,
"sample_count": model.get("sample_count", 0),
})
# Sort by adversarial robustness (descending), then by VI (ascending)
processed_models.sort(key=lambda x: (-x["adversarial_robustness"], x["VI"]))
# Add rankings
for i, model in enumerate(processed_models):
model["rank"] = i + 1
# Find best and worst
best_model = processed_models[0] if processed_models else None
worst_model = processed_models[-1] if processed_models else None
# Find most vulnerable (highest VI)
most_vulnerable = max(processed_models, key=lambda x: x["VI"]) if processed_models else None
# Find most stable (highest RSI)
most_stable = max(processed_models, key=lambda x: x["RSI"]) if processed_models else None
report = {
"benchmark_id": metadata.get("name", "unknown"),
"generated_at": datetime.utcnow().isoformat(),
"metadata": metadata,
"models": processed_models,
"ranking_order": [m["model_name"] for m in processed_models],
"best_model": best_model["model_name"] if best_model else None,
"most_vulnerable_model": most_vulnerable["model_name"] if most_vulnerable else None,
"most_stable_model": most_stable["model_name"] if most_stable else None,
"summary": {
"total_models": len(processed_models),
"average_baseline": sum(m["baseline_robustness"] for m in processed_models) / len(processed_models) if processed_models else 0,
"average_adversarial": sum(m["adversarial_robustness"] for m in processed_models) / len(processed_models) if processed_models else 0,
"average_delta_R": sum(m["delta_R"] for m in processed_models) / len(processed_models) if processed_models else 0,
"average_RSI": sum(m["RSI"] for m in processed_models) / len(processed_models) if processed_models else 0,
"average_VI": sum(m["VI"] for m in processed_models) / len(processed_models) if processed_models else 0,
},
}
return report
def export_benchmark_report(
benchmark_id: str,
format: ExportFormat = ExportFormat.JSON,
include_rankings: bool = True,
include_comparisons: bool = False,
) -> str:
"""
Export benchmark report in specified format.
Args:
benchmark_id: The benchmark identifier
format: Export format (JSON or CSV)
include_rankings: Include rankings in report
include_comparisons: Include pairwise comparisons
Returns:
Formatted report string
"""
# Load benchmark data
benchmark_data = load_benchmark_data(benchmark_id)
if benchmark_data is None:
raise ValueError(f"Benchmark not found: {benchmark_id}")
# Generate report
report = generate_benchmark_report(
benchmark_data,
include_rankings=include_rankings,
include_comparisons=include_comparisons,
)
if format == ExportFormat.JSON:
return json.dumps(report, indent=2)
elif format == ExportFormat.CSV:
return generate_benchmark_csv_report(report)
else:
raise ValueError(f"Unsupported format: {format}")
def generate_benchmark_csv_report(report: Dict[str, Any]) -> str:
"""
Generate CSV report from benchmark report.
Args:
report: Benchmark report dictionary
Returns:
CSV string
"""
output = io.StringIO()
writer = csv.writer(output)
# Header
writer.writerow([
"Rank",
"Model",
"Baseline",
"Adversarial",
"Delta_R",
"RSI",
"VI",
"Samples",
])
# Data rows
for model in report.get("models", []):
writer.writerow([
model.get("rank", ""),
model.get("model_name", ""),
f"{model.get('baseline_robustness', 0):.6f}",
f"{model.get('adversarial_robustness', 0):.6f}",
f"{model.get('delta_R', 0):.6f}",
f"{model.get('RSI', 0):.6f}",
f"{model.get('VI', 0):.6f}",
model.get("sample_count", ""),
])
# Summary rows
writer.writerow([])
writer.writerow(["Summary"])
summary = report.get("summary", {})
writer.writerow(["Total Models", summary.get("total_models", 0)])
writer.writerow(["Average Baseline", f"{summary.get('average_baseline', 0):.6f}"])
writer.writerow(["Average Adversarial", f"{summary.get('average_adversarial', 0):.6f}"])
writer.writerow(["Average Delta_R", f"{summary.get('average_delta_R', 0):.6f}"])
writer.writerow(["Average RSI", f"{summary.get('average_RSI', 0):.6f}"])
writer.writerow(["Average VI", f"{summary.get('average_VI', 0):.6f}"])
writer.writerow([])
writer.writerow(["Best Model", report.get("best_model", "N/A")])
writer.writerow(["Most Vulnerable", report.get("most_vulnerable_model", "N/A")])
writer.writerow(["Most Stable", report.get("most_stable_model", "N/A")])
return output.getvalue()
def save_benchmark_report(
benchmark_id: str,
report: Dict[str, Any],
output_dir: str = "reports",
) -> str:
"""
Save benchmark report to file.
Args:
benchmark_id: The benchmark identifier
report: Report dictionary
output_dir: Output directory
Returns:
Path to saved file
"""
import os
from pathlib import Path
# Create output directory
output_path = Path(output_dir)
output_path.mkdir(parents=True, exist_ok=True)
# Generate filename
filename = f"benchmark_{benchmark_id}.json"
filepath = output_path / filename
# Write file
with open(filepath, "w") as f:
json.dump(report, f, indent=2)
return str(filepath)
|