File size: 7,786 Bytes
a361db3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
"""
Benchmark script: Compare all approaches on test audio files.

Runs all three approaches on each input file, collects metrics,
and produces a CSV comparison report.

Usage:
  uv run python scripts/benchmark.py --data-dir data --output-dir benchmark_results
"""

import argparse
import sys
import csv
import json
import logging
import time
from pathlib import Path
from datetime import datetime
from approaches import list_approaches, get_approach

logging.basicConfig(
    level=logging.INFO,
    format="%(asctime)s [%(levelname)s] %(message)s",
    datefmt="%H:%M:%S",
)
log = logging.getLogger("benchmark")

# Ensure project root is importable when script is run directly
sys.path.insert(0, str(Path(__file__).resolve().parent.parent))



class BenchmarkRunner:
    """Run benchmark across all approaches and audio files."""

    def __init__(self, data_dir: str, output_dir: str):
        self.data_dir = Path(data_dir)
        self.output_dir = Path(output_dir)
        self.output_dir.mkdir(parents=True, exist_ok=True)

        # Results storage
        self.results = []

    def find_test_files(self) -> list:
        """Find all WAV files in data directory."""
        wav_files = sorted(self.data_dir.glob("*.wav"))
        if not wav_files:
            log.warning(f"No WAV files found in {self.data_dir}")
        return wav_files

    def run_benchmark(self, whisper_model: str = "base"):
        """Run all approaches on all test files."""
        test_files = self.find_test_files()

        if not test_files:
            log.error(f"No test files found in {self.data_dir}")
            return

        log.info("="*70)
        log.info(f"BENCHMARK: {len(list_approaches())} approaches × {len(test_files)} files")
        log.info("="*70)

        for input_file in test_files:
            log.info(f"\n{'='*70}")
            log.info(f"File: {input_file.name}")
            log.info(f"{'='*70}")

            for approach_name in list_approaches():
                log.info(f"\n  Testing approach: {approach_name}")
                log.info("-"*70)

                try:
                    result = self._run_approach(
                        approach_name,
                        input_file,
                        whisper_model,
                    )
                    self.results.append(result)

                except Exception as e:
                    log.error(f"  FAILED: {e}")
                    result = {
                        "timestamp": datetime.now().isoformat(),
                        "input_file": input_file.name,
                        "approach": approach_name,
                        "status": "FAILED",
                        "error": str(e),
                    }
                    self.results.append(result)

        log.info(f"\n{'='*70}")
        log.info("BENCHMARK COMPLETE")
        log.info(f"{'='*70}")

        # Save results
        self._save_results()
        self._print_summary()

    def _run_approach(self, approach_name: str, input_file: Path, whisper_model: str):
        """Run single approach on single file."""

        # Create output directory for this run
        output_subdir = self.output_dir / approach_name / input_file.stem
        output_subdir.mkdir(parents=True, exist_ok=True)

        # Initialize approach
        approach_class = get_approach(approach_name)
        approach = approach_class()

        # Run
        start_time = time.time()
        pipeline_output = approach.run(
            input_file=str(input_file),
            output_dir=str(output_subdir),
            whisper_model=whisper_model,
        )
        execution_time = time.time() - start_time

        result = {
            "timestamp": datetime.now().isoformat(),
            "input_file": input_file.name,
            "input_size_mb": input_file.stat().st_size / (1024*1024),
            "approach": approach_name,
            "status": "SUCCESS",
            "duration_seconds": pipeline_output.duration_seconds,
            "execution_time_seconds": execution_time,
            "samples_per_second": (pipeline_output.duration_seconds / execution_time)
                                 if execution_time > 0 else 0,
            "n_speakers": pipeline_output.n_speakers,
            "talker_of_interest": pipeline_output.talker_of_interest,
            "separation_method": pipeline_output.separation_method,
            "doa_method": pipeline_output.doa_method,
            "gender_method": pipeline_output.gender_method,
            "asr_model": pipeline_output.asr_model,
            "output_dir": str(output_subdir),
        }

        # Log metrics
        log.info("    Status: SUCCESS")
        log.info(f"    Execution time: {execution_time:.2f}s")
        log.info(f"    Speakers: {pipeline_output.n_speakers}")
        log.info(f"    ToI: Speaker {pipeline_output.talker_of_interest}")
        log.info(f"    Output: {output_subdir}")

        return result

    def _save_results(self):
        """Save results to CSV and JSON."""

        # Save CSV
        csv_path = self.output_dir / "benchmark_results.csv"
        if self.results:
            fieldnames = self.results[0].keys()
            with open(csv_path, 'w', newline='') as f:
                writer = csv.DictWriter(f, fieldnames=fieldnames)
                writer.writeheader()
                writer.writerows(self.results)
            log.info(f"\nSaved: {csv_path}")

        # Save JSON
        json_path = self.output_dir / "benchmark_results.json"
        with open(json_path, 'w') as f:
            json.dump(self.results, f, indent=2)
        log.info(f"Saved: {json_path}")

    def _print_summary(self):
        """Print summary statistics."""

        if not self.results:
            return

        log.info("\n" + "="*70)
        log.info("SUMMARY")
        log.info("="*70)

        # Group by approach
        by_approach = {}
        for result in self.results:
            approach = result.get("approach")
            if approach not in by_approach:
                by_approach[approach] = []
            by_approach[approach].append(result)

        # Print stats per approach
        for approach, runs in sorted(by_approach.items()):
            successful = [r for r in runs if r.get("status") == "SUCCESS"]
            failed = [r for r in runs if r.get("status") == "FAILED"]

            log.info(f"\nApproach: {approach}")
            log.info(f"  Successful: {len(successful)}/{len(runs)}")

            if successful:
                avg_exec_time = sum(r["execution_time_seconds"] for r in successful) / len(successful)
                avg_speedup = sum(r.get("samples_per_second", 0) for r in successful) / len(successful)

                log.info(f"  Avg execution time: {avg_exec_time:.2f}s")
                log.info(f"  Avg speedup (samples/s): {avg_speedup:.1f}x")

            if failed:
                log.info(f"  Failed runs: {len(failed)}")


def main():
    parser = argparse.ArgumentParser(description="Benchmark all approaches")
    parser.add_argument("--data-dir", default="data", help="Directory with test WAV files")
    parser.add_argument("--output-dir", default="benchmark_results", help="Output directory")
    parser.add_argument("-w", "--whisper-model", default="base", help="Whisper model")
    parser.add_argument("-v", "--verbose", action="store_true")

    args = parser.parse_args()

    if args.verbose:
        logging.getLogger().setLevel(logging.DEBUG)

    data_path = Path(args.data_dir)
    if not data_path.exists():
        log.error(f"Data directory not found: {data_path}")
        return 1

    runner = BenchmarkRunner(args.data_dir, args.output_dir)
    runner.run_benchmark(args.whisper_model)

    return 0


if __name__ == "__main__":
    sys.exit(main())