Repository navigation
Expand file tree
/
Copy pathmuscriptor_python_warm_bench.py
More file actions
262 lines (227 loc) · 10.3 KB
/
Copy pathmuscriptor_python_warm_bench.py
File metadata and controls
262 lines (227 loc) · 10.3 KB
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
#!/usr/bin/env python3
from __future__ import annotations
import argparse
import dataclasses
import json
import sys
import time
from pathlib import Path
from typing import Any
import torch
REPO_ROOT = Path(__file__).resolve().parents[2]
REFERENCE_ROOT = REPO_ROOT / "reference" / "muscriptor"
sys.path.insert(0, str(REFERENCE_ROOT))
from muscriptor.events import NoteEndEvent, NoteStartEvent, ProgressEvent # noqa: E402
import muscriptor.accelerator # noqa: E402
from muscriptor.tokenizer.mt3 import resolve_instrument_names # noqa: E402
from muscriptor.transcription_model import TranscriptionModel # noqa: E402
def resolve_path(path: str | Path) -> Path:
value = Path(path)
return value if value.is_absolute() else REPO_ROOT / value
def parse_csv_paths(value: str, fallback: Path) -> list[Path]:
return [Path(item) for item in value.split(",") if item] if value else [fallback]
def parse_csv_keep_empty(value: str) -> list[str]:
if value.strip().startswith("["):
parsed = json.loads(value)
if not isinstance(parsed, list) or not all(isinstance(item, str) for item in parsed):
raise ValueError("--instruments-sequence must be a JSON string array")
return parsed
return value.split(",") if value else []
def parse_sequence(value: str) -> list[Any]:
if not value:
return []
if value.strip().startswith("["):
parsed = json.loads(value)
if not isinstance(parsed, list):
raise ValueError("sequence arguments must be JSON arrays")
return parsed
return value.split(",")
def sequence_value(values: list[Any], index: int, fallback: Any) -> Any:
return values[index] if index < len(values) else fallback
def parse_bool(value: Any) -> bool:
if isinstance(value, bool):
return value
normalized = str(value).strip().lower()
if normalized in {"1", "true", "yes", "on"}:
return True
if normalized in {"0", "false", "no", "off"}:
return False
raise ValueError(f"invalid bool value: {value}")
def optional_positive_int(value: Any) -> int | None:
parsed = int(value)
return parsed if parsed > 0 else None
def event_to_dict(event: NoteStartEvent | NoteEndEvent) -> dict[str, Any]:
if isinstance(event, NoteStartEvent):
return {"type": "start", **dataclasses.asdict(event)}
return {
"type": "end",
"end_time": event.end_time,
"start_event_index": event.start_event_index,
}
def normalized_instruments(value: str) -> list[str] | None:
if not value:
return None
return resolve_instrument_names([item for item in value.split(",") if item.strip()])
def transcribe_json(
model: TranscriptionModel,
audio: Path,
instruments: str,
*,
use_sampling: bool,
temperature: float,
cfg_coef: float,
batch_size: int | None,
beam_size: int,
prelude_forcing: bool,
) -> str:
events = [
event_to_dict(event)
for event in model.transcribe(
audio=audio,
instruments=normalized_instruments(instruments),
use_sampling=use_sampling,
temperature=temperature,
cfg_coef=cfg_coef,
batch_size=batch_size,
beam_size=beam_size,
prelude_forcing=prelude_forcing,
)
if not isinstance(event, ProgressEvent)
]
return json.dumps(events, separators=(",", ":"))
def main() -> int:
parser = argparse.ArgumentParser(description="Python reference MuScriptor warmbench.")
parser.add_argument("--model", default="models/muscriptor-small/model.safetensors")
parser.add_argument("--audio", default="build/logs/muscriptor/input/headache_by_lost_deposit_1min_16k.wav")
parser.add_argument("--warmup-audio", default="")
parser.add_argument("--audio-sequence", default="")
parser.add_argument("--backend", choices=["cpu", "cuda"], default="cpu")
parser.add_argument("--device", type=int, default=0)
parser.add_argument("--threads", type=int, default=8)
parser.add_argument("--warmup", type=int, default=1)
parser.add_argument("--iterations", type=int, default=1)
parser.add_argument("--instruments", default="")
parser.add_argument("--instruments-sequence", default="")
parser.add_argument("--use-sampling", default="false")
parser.add_argument("--use-sampling-sequence", default="")
parser.add_argument("--temperature", type=float, default=1.0)
parser.add_argument("--temperature-sequence", default="")
parser.add_argument("--cfg-coef", type=float, default=1.0)
parser.add_argument("--cfg-coef-sequence", default="")
parser.add_argument("--batch-size", type=int, default=0)
parser.add_argument("--batch-size-sequence", default="")
parser.add_argument("--beam-size", type=int, default=1)
parser.add_argument("--beam-size-sequence", default="")
parser.add_argument("--prelude-forcing", default="true")
parser.add_argument("--prelude-forcing-sequence", default="")
parser.add_argument("--seed", type=int, default=1234)
parser.add_argument("--seed-sequence", default="")
parser.add_argument("--timing-file", default="")
parser.add_argument("--disable-cuda-autocast", action="store_true")
args = parser.parse_args()
torch.set_num_threads(max(1, args.threads))
if args.backend == "cuda":
if not torch.cuda.is_available():
raise RuntimeError("CUDA is not available")
device = f"cuda:{args.device}"
else:
device = "cpu"
model_path = resolve_path(args.model)
model = TranscriptionModel.load_model(weights_path=model_path, device=device)
if args.disable_cuda_autocast:
model._model.autocast.enabled = False
warmup_audio = resolve_path(args.warmup_audio) if args.warmup_audio else resolve_path(args.audio)
for _ in range(args.warmup):
torch.manual_seed(args.seed)
transcribe_json(
model,
warmup_audio,
args.instruments,
use_sampling=parse_bool(args.use_sampling),
temperature=args.temperature,
cfg_coef=args.cfg_coef,
batch_size=optional_positive_int(args.batch_size),
beam_size=args.beam_size,
prelude_forcing=parse_bool(args.prelude_forcing),
)
request_paths = parse_csv_paths(args.audio_sequence, Path(args.audio))
instrument_values = parse_csv_keep_empty(args.instruments_sequence)
use_sampling_values = parse_sequence(args.use_sampling_sequence)
temperature_values = parse_sequence(args.temperature_sequence)
cfg_coef_values = parse_sequence(args.cfg_coef_sequence)
batch_size_values = parse_sequence(args.batch_size_sequence)
beam_size_values = parse_sequence(args.beam_size_sequence)
prelude_forcing_values = parse_sequence(args.prelude_forcing_sequence)
seed_values = parse_sequence(args.seed_sequence)
timing_lines = [
f"muscriptor.model_root {model_path}",
f"muscriptor.backend {args.backend}",
]
steps = []
for request_index, audio in enumerate(request_paths):
audio_path = resolve_path(audio)
instruments = sequence_value(instrument_values, request_index, args.instruments)
use_sampling = parse_bool(sequence_value(use_sampling_values, request_index, args.use_sampling))
temperature = float(sequence_value(temperature_values, request_index, args.temperature))
cfg_coef = float(sequence_value(cfg_coef_values, request_index, args.cfg_coef))
batch_size = optional_positive_int(sequence_value(batch_size_values, request_index, args.batch_size))
beam_size = int(sequence_value(beam_size_values, request_index, args.beam_size))
prelude_forcing = parse_bool(sequence_value(prelude_forcing_values, request_index, args.prelude_forcing))
seed = int(sequence_value(seed_values, request_index, args.seed))
text_output = ""
total_ms = 0.0
for _ in range(args.iterations):
torch.manual_seed(seed)
muscriptor.accelerator.synchronize()
started = time.perf_counter()
text_output = transcribe_json(
model,
audio_path,
instruments,
use_sampling=use_sampling,
temperature=temperature,
cfg_coef=cfg_coef,
batch_size=batch_size,
beam_size=beam_size,
prelude_forcing=prelude_forcing,
)
muscriptor.accelerator.synchronize()
total_ms += (time.perf_counter() - started) * 1000.0
wall_ms = total_ms / args.iterations
print(f"average[{request_index}]")
print(f"muscriptor.wall_ms={wall_ms}")
timing_lines.append(f"muscriptor.request{request_index}.wall_ms {wall_ms:.6f}")
timing_lines.append(f"muscriptor.request{request_index}.instruments {instruments}")
timing_lines.append(f"muscriptor.request{request_index}.use_sampling {int(use_sampling)}")
timing_lines.append(f"muscriptor.request{request_index}.temperature {temperature}")
timing_lines.append(f"muscriptor.request{request_index}.cfg_coef {cfg_coef}")
timing_lines.append(f"muscriptor.request{request_index}.batch_size {batch_size or 0}")
timing_lines.append(f"muscriptor.request{request_index}.beam_size {beam_size}")
timing_lines.append(f"muscriptor.request{request_index}.prelude_forcing {int(prelude_forcing)}")
timing_lines.append(f"muscriptor.request{request_index}.seed {seed}")
steps.append({
"request_index": request_index,
"audio": str(audio),
"instruments": instruments,
"use_sampling": use_sampling,
"temperature": temperature,
"cfg_coef": cfg_coef,
"batch_size": batch_size,
"beam_size": beam_size,
"prelude_forcing": prelude_forcing,
"seed": seed,
"text_output": text_output,
"word_timestamps": [],
"metrics": {"wall_ms": wall_ms},
})
if args.timing_file:
timing_path = Path(args.timing_file)
timing_path.parent.mkdir(parents=True, exist_ok=True)
timing_path.write_text("\n".join(timing_lines) + "\n", encoding="utf-8")
print("summary_json=" + json.dumps(
{"family": "muscriptor", "backend": args.backend, "sequence_steps": steps},
separators=(",", ":"),
))
return 0
if __name__ == "__main__":
raise SystemExit(main())