Update tools/api.py
Browse files- tools/api.py +440 -440
tools/api.py
CHANGED
@@ -1,440 +1,440 @@
|
|
1 |
-
import base64
|
2 |
-
import io
|
3 |
-
import json
|
4 |
-
import queue
|
5 |
-
import random
|
6 |
-
import sys
|
7 |
-
import traceback
|
8 |
-
import wave
|
9 |
-
from argparse import ArgumentParser
|
10 |
-
from http import HTTPStatus
|
11 |
-
from pathlib import Path
|
12 |
-
from typing import Annotated, Any, Literal, Optional
|
13 |
-
|
14 |
-
import numpy as np
|
15 |
-
import ormsgpack
|
16 |
-
import pyrootutils
|
17 |
-
import soundfile as sf
|
18 |
-
import torch
|
19 |
-
import torchaudio
|
20 |
-
from baize.datastructures import ContentType
|
21 |
-
from kui.asgi import (
|
22 |
-
Body,
|
23 |
-
FactoryClass,
|
24 |
-
HTTPException,
|
25 |
-
HttpRequest,
|
26 |
-
HttpView,
|
27 |
-
JSONResponse,
|
28 |
-
Kui,
|
29 |
-
OpenAPI,
|
30 |
-
StreamResponse,
|
31 |
-
)
|
32 |
-
from kui.asgi.routing import MultimethodRoutes
|
33 |
-
from loguru import logger
|
34 |
-
from pydantic import BaseModel, Field, conint
|
35 |
-
|
36 |
-
pyrootutils.setup_root(__file__, indicator=".project-root", pythonpath=True)
|
37 |
-
|
38 |
-
# from fish_speech.models.vqgan.lit_module import VQGAN
|
39 |
-
from fish_speech.models.vqgan.modules.firefly import FireflyArchitecture
|
40 |
-
from fish_speech.text.chn_text_norm.text import Text as ChnNormedText
|
41 |
-
from fish_speech.utils import autocast_exclude_mps
|
42 |
-
from tools.commons import ServeReferenceAudio, ServeTTSRequest
|
43 |
-
from tools.file import AUDIO_EXTENSIONS, audio_to_bytes, list_files, read_ref_text
|
44 |
-
from tools.llama.generate import (
|
45 |
-
GenerateRequest,
|
46 |
-
GenerateResponse,
|
47 |
-
WrappedGenerateResponse,
|
48 |
-
launch_thread_safe_queue,
|
49 |
-
)
|
50 |
-
from tools.vqgan.inference import load_model as load_decoder_model
|
51 |
-
|
52 |
-
|
53 |
-
def wav_chunk_header(sample_rate=44100, bit_depth=16, channels=1):
|
54 |
-
buffer = io.BytesIO()
|
55 |
-
|
56 |
-
with wave.open(buffer, "wb") as wav_file:
|
57 |
-
wav_file.setnchannels(channels)
|
58 |
-
wav_file.setsampwidth(bit_depth // 8)
|
59 |
-
wav_file.setframerate(sample_rate)
|
60 |
-
|
61 |
-
wav_header_bytes = buffer.getvalue()
|
62 |
-
buffer.close()
|
63 |
-
return wav_header_bytes
|
64 |
-
|
65 |
-
|
66 |
-
# Define utils for web server
|
67 |
-
async def http_execption_handler(exc: HTTPException):
|
68 |
-
return JSONResponse(
|
69 |
-
dict(
|
70 |
-
statusCode=exc.status_code,
|
71 |
-
message=exc.content,
|
72 |
-
error=HTTPStatus(exc.status_code).phrase,
|
73 |
-
),
|
74 |
-
exc.status_code,
|
75 |
-
exc.headers,
|
76 |
-
)
|
77 |
-
|
78 |
-
|
79 |
-
async def other_exception_handler(exc: "Exception"):
|
80 |
-
traceback.print_exc()
|
81 |
-
|
82 |
-
status = HTTPStatus.INTERNAL_SERVER_ERROR
|
83 |
-
return JSONResponse(
|
84 |
-
dict(statusCode=status, message=str(exc), error=status.phrase),
|
85 |
-
status,
|
86 |
-
)
|
87 |
-
|
88 |
-
|
89 |
-
def load_audio(reference_audio, sr):
|
90 |
-
if len(reference_audio) > 255 or not Path(reference_audio).exists():
|
91 |
-
audio_data = reference_audio
|
92 |
-
reference_audio = io.BytesIO(audio_data)
|
93 |
-
|
94 |
-
waveform, original_sr = torchaudio.load(
|
95 |
-
reference_audio, backend="sox" if sys.platform == "linux" else "soundfile"
|
96 |
-
)
|
97 |
-
|
98 |
-
if waveform.shape[0] > 1:
|
99 |
-
waveform = torch.mean(waveform, dim=0, keepdim=True)
|
100 |
-
|
101 |
-
if original_sr != sr:
|
102 |
-
resampler = torchaudio.transforms.Resample(orig_freq=original_sr, new_freq=sr)
|
103 |
-
waveform = resampler(waveform)
|
104 |
-
|
105 |
-
audio = waveform.squeeze().numpy()
|
106 |
-
return audio
|
107 |
-
|
108 |
-
|
109 |
-
def encode_reference(*, decoder_model, reference_audio, enable_reference_audio):
|
110 |
-
if enable_reference_audio and reference_audio is not None:
|
111 |
-
# Load audios, and prepare basic info here
|
112 |
-
reference_audio_content = load_audio(
|
113 |
-
reference_audio, decoder_model.spec_transform.sample_rate
|
114 |
-
)
|
115 |
-
|
116 |
-
audios = torch.from_numpy(reference_audio_content).to(decoder_model.device)[
|
117 |
-
None, None, :
|
118 |
-
]
|
119 |
-
audio_lengths = torch.tensor(
|
120 |
-
[audios.shape[2]], device=decoder_model.device, dtype=torch.long
|
121 |
-
)
|
122 |
-
logger.info(
|
123 |
-
f"Loaded audio with {audios.shape[2] / decoder_model.spec_transform.sample_rate:.2f} seconds"
|
124 |
-
)
|
125 |
-
|
126 |
-
# VQ Encoder
|
127 |
-
if isinstance(decoder_model, FireflyArchitecture):
|
128 |
-
prompt_tokens = decoder_model.encode(audios, audio_lengths)[0][0]
|
129 |
-
|
130 |
-
logger.info(f"Encoded prompt: {prompt_tokens.shape}")
|
131 |
-
else:
|
132 |
-
prompt_tokens = None
|
133 |
-
logger.info("No reference audio provided")
|
134 |
-
|
135 |
-
return prompt_tokens
|
136 |
-
|
137 |
-
|
138 |
-
def decode_vq_tokens(
|
139 |
-
*,
|
140 |
-
decoder_model,
|
141 |
-
codes,
|
142 |
-
):
|
143 |
-
feature_lengths = torch.tensor([codes.shape[1]], device=decoder_model.device)
|
144 |
-
logger.info(f"VQ features: {codes.shape}")
|
145 |
-
|
146 |
-
if isinstance(decoder_model, FireflyArchitecture):
|
147 |
-
# VQGAN Inference
|
148 |
-
return decoder_model.decode(
|
149 |
-
indices=codes[None],
|
150 |
-
feature_lengths=feature_lengths,
|
151 |
-
)[0].squeeze()
|
152 |
-
|
153 |
-
raise ValueError(f"Unknown model type: {type(decoder_model)}")
|
154 |
-
|
155 |
-
|
156 |
-
routes = MultimethodRoutes(base_class=HttpView)
|
157 |
-
|
158 |
-
|
159 |
-
def get_content_type(audio_format):
|
160 |
-
if audio_format == "wav":
|
161 |
-
return "audio/wav"
|
162 |
-
elif audio_format == "flac":
|
163 |
-
return "audio/flac"
|
164 |
-
elif audio_format == "mp3":
|
165 |
-
return "audio/mpeg"
|
166 |
-
else:
|
167 |
-
return "application/octet-stream"
|
168 |
-
|
169 |
-
|
170 |
-
@torch.inference_mode()
|
171 |
-
def inference(req: ServeTTSRequest):
|
172 |
-
|
173 |
-
idstr: str | None = req.reference_id
|
174 |
-
if idstr is not None:
|
175 |
-
ref_folder = Path("references") / idstr
|
176 |
-
ref_folder.mkdir(parents=True, exist_ok=True)
|
177 |
-
ref_audios = list_files(
|
178 |
-
ref_folder, AUDIO_EXTENSIONS, recursive=True, sort=False
|
179 |
-
)
|
180 |
-
prompt_tokens = [
|
181 |
-
encode_reference(
|
182 |
-
decoder_model=decoder_model,
|
183 |
-
reference_audio=audio_to_bytes(str(ref_audio)),
|
184 |
-
enable_reference_audio=True,
|
185 |
-
)
|
186 |
-
for ref_audio in ref_audios
|
187 |
-
]
|
188 |
-
prompt_texts = [
|
189 |
-
read_ref_text(str(ref_audio.with_suffix(".lab")))
|
190 |
-
for ref_audio in ref_audios
|
191 |
-
]
|
192 |
-
|
193 |
-
else:
|
194 |
-
# Parse reference audio aka prompt
|
195 |
-
refs = req.references
|
196 |
-
if refs is None:
|
197 |
-
refs = []
|
198 |
-
prompt_tokens = [
|
199 |
-
encode_reference(
|
200 |
-
decoder_model=decoder_model,
|
201 |
-
reference_audio=ref.audio,
|
202 |
-
enable_reference_audio=True,
|
203 |
-
)
|
204 |
-
for ref in refs
|
205 |
-
]
|
206 |
-
prompt_texts = [ref.text for ref in refs]
|
207 |
-
|
208 |
-
# LLAMA Inference
|
209 |
-
request = dict(
|
210 |
-
device=decoder_model.device,
|
211 |
-
max_new_tokens=req.max_new_tokens,
|
212 |
-
text=(
|
213 |
-
req.text
|
214 |
-
if not req.normalize
|
215 |
-
else ChnNormedText(raw_text=req.text).normalize()
|
216 |
-
),
|
217 |
-
top_p=req.top_p,
|
218 |
-
repetition_penalty=req.repetition_penalty,
|
219 |
-
temperature=req.temperature,
|
220 |
-
compile=args.compile,
|
221 |
-
iterative_prompt=req.chunk_length > 0,
|
222 |
-
chunk_length=req.chunk_length,
|
223 |
-
max_length=2048,
|
224 |
-
prompt_tokens=prompt_tokens,
|
225 |
-
prompt_text=prompt_texts,
|
226 |
-
)
|
227 |
-
|
228 |
-
response_queue = queue.Queue()
|
229 |
-
llama_queue.put(
|
230 |
-
GenerateRequest(
|
231 |
-
request=request,
|
232 |
-
response_queue=response_queue,
|
233 |
-
)
|
234 |
-
)
|
235 |
-
|
236 |
-
if req.streaming:
|
237 |
-
yield wav_chunk_header()
|
238 |
-
|
239 |
-
segments = []
|
240 |
-
while True:
|
241 |
-
result: WrappedGenerateResponse = response_queue.get()
|
242 |
-
if result.status == "error":
|
243 |
-
raise result.response
|
244 |
-
break
|
245 |
-
|
246 |
-
result: GenerateResponse = result.response
|
247 |
-
if result.action == "next":
|
248 |
-
break
|
249 |
-
|
250 |
-
with autocast_exclude_mps(
|
251 |
-
device_type=decoder_model.device.type, dtype=args.precision
|
252 |
-
):
|
253 |
-
fake_audios = decode_vq_tokens(
|
254 |
-
decoder_model=decoder_model,
|
255 |
-
codes=result.codes,
|
256 |
-
)
|
257 |
-
|
258 |
-
fake_audios = fake_audios.float().cpu().numpy()
|
259 |
-
|
260 |
-
if req.streaming:
|
261 |
-
yield (fake_audios * 32768).astype(np.int16).tobytes()
|
262 |
-
else:
|
263 |
-
segments.append(fake_audios)
|
264 |
-
|
265 |
-
if req.streaming:
|
266 |
-
return
|
267 |
-
|
268 |
-
if len(segments) == 0:
|
269 |
-
raise HTTPException(
|
270 |
-
HTTPStatus.INTERNAL_SERVER_ERROR,
|
271 |
-
content="No audio generated, please check the input text.",
|
272 |
-
)
|
273 |
-
|
274 |
-
fake_audios = np.concatenate(segments, axis=0)
|
275 |
-
yield fake_audios
|
276 |
-
|
277 |
-
|
278 |
-
async def inference_async(req: ServeTTSRequest):
|
279 |
-
for chunk in inference(req):
|
280 |
-
yield chunk
|
281 |
-
|
282 |
-
|
283 |
-
async def buffer_to_async_generator(buffer):
|
284 |
-
yield buffer
|
285 |
-
|
286 |
-
|
287 |
-
@routes.http.post("/v1/tts")
|
288 |
-
async def api_invoke_model(
|
289 |
-
req: Annotated[ServeTTSRequest, Body(exclusive=True)],
|
290 |
-
):
|
291 |
-
"""
|
292 |
-
Invoke model and generate audio
|
293 |
-
"""
|
294 |
-
|
295 |
-
if args.max_text_length > 0 and len(req.text) > args.max_text_length:
|
296 |
-
raise HTTPException(
|
297 |
-
HTTPStatus.BAD_REQUEST,
|
298 |
-
content=f"Text is too long, max length is {args.max_text_length}",
|
299 |
-
)
|
300 |
-
|
301 |
-
if req.streaming and req.format != "wav":
|
302 |
-
raise HTTPException(
|
303 |
-
HTTPStatus.BAD_REQUEST,
|
304 |
-
content="Streaming only supports WAV format",
|
305 |
-
)
|
306 |
-
|
307 |
-
if req.streaming:
|
308 |
-
return StreamResponse(
|
309 |
-
iterable=inference_async(req),
|
310 |
-
headers={
|
311 |
-
"Content-Disposition": f"attachment; filename=audio.{req.format}",
|
312 |
-
},
|
313 |
-
content_type=get_content_type(req.format),
|
314 |
-
)
|
315 |
-
else:
|
316 |
-
fake_audios = next(inference(req))
|
317 |
-
buffer = io.BytesIO()
|
318 |
-
sf.write(
|
319 |
-
buffer,
|
320 |
-
fake_audios,
|
321 |
-
decoder_model.spec_transform.sample_rate,
|
322 |
-
format=req.format,
|
323 |
-
)
|
324 |
-
|
325 |
-
return StreamResponse(
|
326 |
-
iterable=buffer_to_async_generator(buffer.getvalue()),
|
327 |
-
headers={
|
328 |
-
"Content-Disposition": f"attachment; filename=audio.{req.format}",
|
329 |
-
},
|
330 |
-
content_type=get_content_type(req.format),
|
331 |
-
)
|
332 |
-
|
333 |
-
|
334 |
-
@routes.http.post("/v1/health")
|
335 |
-
async def api_health():
|
336 |
-
"""
|
337 |
-
Health check
|
338 |
-
"""
|
339 |
-
|
340 |
-
return JSONResponse({"status": "ok"})
|
341 |
-
|
342 |
-
|
343 |
-
def parse_args():
|
344 |
-
parser = ArgumentParser()
|
345 |
-
parser.add_argument(
|
346 |
-
"--llama-checkpoint-path",
|
347 |
-
type=str,
|
348 |
-
default="checkpoints/fish-speech-1.4",
|
349 |
-
)
|
350 |
-
parser.add_argument(
|
351 |
-
"--decoder-checkpoint-path",
|
352 |
-
type=str,
|
353 |
-
default="checkpoints/fish-speech-1.4/firefly-gan-vq-fsq-8x1024-21hz-generator.pth",
|
354 |
-
)
|
355 |
-
parser.add_argument("--decoder-config-name", type=str, default="firefly_gan_vq")
|
356 |
-
parser.add_argument("--device", type=str, default="cuda")
|
357 |
-
parser.add_argument("--half", action="store_true")
|
358 |
-
parser.add_argument("--compile", action="store_true")
|
359 |
-
parser.add_argument("--max-text-length", type=int, default=0)
|
360 |
-
parser.add_argument("--listen", type=str, default="127.0.0.1:8080")
|
361 |
-
parser.add_argument("--workers", type=int, default=1)
|
362 |
-
|
363 |
-
return parser.parse_args()
|
364 |
-
|
365 |
-
|
366 |
-
# Define Kui app
|
367 |
-
openapi = OpenAPI(
|
368 |
-
{
|
369 |
-
"title": "Fish Speech API",
|
370 |
-
},
|
371 |
-
).routes
|
372 |
-
|
373 |
-
|
374 |
-
class MsgPackRequest(HttpRequest):
|
375 |
-
async def data(self) -> Annotated[Any, ContentType("application/msgpack")]:
|
376 |
-
if self.content_type == "application/msgpack":
|
377 |
-
return ormsgpack.unpackb(await self.body)
|
378 |
-
|
379 |
-
raise HTTPException(
|
380 |
-
HTTPStatus.UNSUPPORTED_MEDIA_TYPE,
|
381 |
-
headers={"Accept": "application/msgpack"},
|
382 |
-
)
|
383 |
-
|
384 |
-
|
385 |
-
app = Kui(
|
386 |
-
routes=routes + openapi[1:], # Remove the default route
|
387 |
-
exception_handlers={
|
388 |
-
HTTPException: http_execption_handler,
|
389 |
-
Exception: other_exception_handler,
|
390 |
-
},
|
391 |
-
factory_class=FactoryClass(http=MsgPackRequest),
|
392 |
-
cors_config={},
|
393 |
-
)
|
394 |
-
|
395 |
-
|
396 |
-
if __name__ == "__main__":
|
397 |
-
|
398 |
-
import uvicorn
|
399 |
-
|
400 |
-
args = parse_args()
|
401 |
-
args.precision = torch.half if args.half else torch.bfloat16
|
402 |
-
|
403 |
-
logger.info("Loading Llama model...")
|
404 |
-
llama_queue = launch_thread_safe_queue(
|
405 |
-
checkpoint_path=args.llama_checkpoint_path,
|
406 |
-
device=args.device,
|
407 |
-
precision=args.precision,
|
408 |
-
compile=args.compile,
|
409 |
-
)
|
410 |
-
logger.info("Llama model loaded, loading VQ-GAN model...")
|
411 |
-
|
412 |
-
decoder_model = load_decoder_model(
|
413 |
-
config_name=args.decoder_config_name,
|
414 |
-
checkpoint_path=args.decoder_checkpoint_path,
|
415 |
-
device=args.device,
|
416 |
-
)
|
417 |
-
|
418 |
-
logger.info("VQ-GAN model loaded, warming up...")
|
419 |
-
|
420 |
-
# Dry run to check if the model is loaded correctly and avoid the first-time latency
|
421 |
-
list(
|
422 |
-
inference(
|
423 |
-
ServeTTSRequest(
|
424 |
-
text="Hello world.",
|
425 |
-
references=[],
|
426 |
-
reference_id=None,
|
427 |
-
max_new_tokens=1024,
|
428 |
-
chunk_length=200,
|
429 |
-
top_p=0.7,
|
430 |
-
repetition_penalty=1.2,
|
431 |
-
temperature=0.7,
|
432 |
-
emotion=None,
|
433 |
-
format="wav",
|
434 |
-
)
|
435 |
-
)
|
436 |
-
)
|
437 |
-
|
438 |
-
logger.info(f"Warming up done, starting server at http://{args.listen}")
|
439 |
-
host, port = args.listen.split(":")
|
440 |
-
uvicorn.run(app, host=host, port=int(port), workers=args.workers, log_level="info")
|
|
|
1 |
+
import base64
|
2 |
+
import io
|
3 |
+
import json
|
4 |
+
import queue
|
5 |
+
import random
|
6 |
+
import sys
|
7 |
+
import traceback
|
8 |
+
import wave
|
9 |
+
from argparse import ArgumentParser
|
10 |
+
from http import HTTPStatus
|
11 |
+
from pathlib import Path
|
12 |
+
from typing import Annotated, Any, Literal, Optional
|
13 |
+
|
14 |
+
import numpy as np
|
15 |
+
import ormsgpack
|
16 |
+
import pyrootutils
|
17 |
+
import soundfile as sf
|
18 |
+
import torch
|
19 |
+
import torchaudio
|
20 |
+
from baize.datastructures import ContentType
|
21 |
+
from kui.asgi import (
|
22 |
+
Body,
|
23 |
+
FactoryClass,
|
24 |
+
HTTPException,
|
25 |
+
HttpRequest,
|
26 |
+
HttpView,
|
27 |
+
JSONResponse,
|
28 |
+
Kui,
|
29 |
+
OpenAPI,
|
30 |
+
StreamResponse,
|
31 |
+
)
|
32 |
+
from kui.asgi.routing import MultimethodRoutes
|
33 |
+
from loguru import logger
|
34 |
+
from pydantic import BaseModel, Field, conint
|
35 |
+
|
36 |
+
pyrootutils.setup_root(__file__, indicator=".project-root", pythonpath=True)
|
37 |
+
|
38 |
+
# from fish_speech.models.vqgan.lit_module import VQGAN
|
39 |
+
from fish_speech.models.vqgan.modules.firefly import FireflyArchitecture
|
40 |
+
from fish_speech.text.chn_text_norm.text import Text as ChnNormedText
|
41 |
+
from fish_speech.utils import autocast_exclude_mps
|
42 |
+
from tools.commons import ServeReferenceAudio, ServeTTSRequest
|
43 |
+
from tools.file import AUDIO_EXTENSIONS, audio_to_bytes, list_files, read_ref_text
|
44 |
+
from tools.llama.generate import (
|
45 |
+
GenerateRequest,
|
46 |
+
GenerateResponse,
|
47 |
+
WrappedGenerateResponse,
|
48 |
+
launch_thread_safe_queue,
|
49 |
+
)
|
50 |
+
from tools.vqgan.inference import load_model as load_decoder_model
|
51 |
+
|
52 |
+
|
53 |
+
def wav_chunk_header(sample_rate=44100, bit_depth=16, channels=1):
|
54 |
+
buffer = io.BytesIO()
|
55 |
+
|
56 |
+
with wave.open(buffer, "wb") as wav_file:
|
57 |
+
wav_file.setnchannels(channels)
|
58 |
+
wav_file.setsampwidth(bit_depth // 8)
|
59 |
+
wav_file.setframerate(sample_rate)
|
60 |
+
|
61 |
+
wav_header_bytes = buffer.getvalue()
|
62 |
+
buffer.close()
|
63 |
+
return wav_header_bytes
|
64 |
+
|
65 |
+
|
66 |
+
# Define utils for web server
|
67 |
+
async def http_execption_handler(exc: HTTPException):
|
68 |
+
return JSONResponse(
|
69 |
+
dict(
|
70 |
+
statusCode=exc.status_code,
|
71 |
+
message=exc.content,
|
72 |
+
error=HTTPStatus(exc.status_code).phrase,
|
73 |
+
),
|
74 |
+
exc.status_code,
|
75 |
+
exc.headers,
|
76 |
+
)
|
77 |
+
|
78 |
+
|
79 |
+
async def other_exception_handler(exc: "Exception"):
|
80 |
+
traceback.print_exc()
|
81 |
+
|
82 |
+
status = HTTPStatus.INTERNAL_SERVER_ERROR
|
83 |
+
return JSONResponse(
|
84 |
+
dict(statusCode=status, message=str(exc), error=status.phrase),
|
85 |
+
status,
|
86 |
+
)
|
87 |
+
|
88 |
+
|
89 |
+
def load_audio(reference_audio, sr):
|
90 |
+
if len(reference_audio) > 255 or not Path(reference_audio).exists():
|
91 |
+
audio_data = reference_audio
|
92 |
+
reference_audio = io.BytesIO(audio_data)
|
93 |
+
|
94 |
+
waveform, original_sr = torchaudio.load(
|
95 |
+
reference_audio, backend="sox" if sys.platform == "linux" else "soundfile"
|
96 |
+
)
|
97 |
+
|
98 |
+
if waveform.shape[0] > 1:
|
99 |
+
waveform = torch.mean(waveform, dim=0, keepdim=True)
|
100 |
+
|
101 |
+
if original_sr != sr:
|
102 |
+
resampler = torchaudio.transforms.Resample(orig_freq=original_sr, new_freq=sr)
|
103 |
+
waveform = resampler(waveform)
|
104 |
+
|
105 |
+
audio = waveform.squeeze().numpy()
|
106 |
+
return audio
|
107 |
+
|
108 |
+
|
109 |
+
def encode_reference(*, decoder_model, reference_audio, enable_reference_audio):
|
110 |
+
if enable_reference_audio and reference_audio is not None:
|
111 |
+
# Load audios, and prepare basic info here
|
112 |
+
reference_audio_content = load_audio(
|
113 |
+
reference_audio, decoder_model.spec_transform.sample_rate
|
114 |
+
)
|
115 |
+
|
116 |
+
audios = torch.from_numpy(reference_audio_content).to(decoder_model.device)[
|
117 |
+
None, None, :
|
118 |
+
]
|
119 |
+
audio_lengths = torch.tensor(
|
120 |
+
[audios.shape[2]], device=decoder_model.device, dtype=torch.long
|
121 |
+
)
|
122 |
+
logger.info(
|
123 |
+
f"Loaded audio with {audios.shape[2] / decoder_model.spec_transform.sample_rate:.2f} seconds"
|
124 |
+
)
|
125 |
+
|
126 |
+
# VQ Encoder
|
127 |
+
if isinstance(decoder_model, FireflyArchitecture):
|
128 |
+
prompt_tokens = decoder_model.encode(audios, audio_lengths)[0][0]
|
129 |
+
|
130 |
+
logger.info(f"Encoded prompt: {prompt_tokens.shape}")
|
131 |
+
else:
|
132 |
+
prompt_tokens = None
|
133 |
+
logger.info("No reference audio provided")
|
134 |
+
|
135 |
+
return prompt_tokens
|
136 |
+
|
137 |
+
|
138 |
+
def decode_vq_tokens(
|
139 |
+
*,
|
140 |
+
decoder_model,
|
141 |
+
codes,
|
142 |
+
):
|
143 |
+
feature_lengths = torch.tensor([codes.shape[1]], device=decoder_model.device)
|
144 |
+
logger.info(f"VQ features: {codes.shape}")
|
145 |
+
|
146 |
+
if isinstance(decoder_model, FireflyArchitecture):
|
147 |
+
# VQGAN Inference
|
148 |
+
return decoder_model.decode(
|
149 |
+
indices=codes[None],
|
150 |
+
feature_lengths=feature_lengths,
|
151 |
+
)[0].squeeze()
|
152 |
+
|
153 |
+
raise ValueError(f"Unknown model type: {type(decoder_model)}")
|
154 |
+
|
155 |
+
|
156 |
+
routes = MultimethodRoutes(base_class=HttpView)
|
157 |
+
|
158 |
+
|
159 |
+
def get_content_type(audio_format):
|
160 |
+
if audio_format == "wav":
|
161 |
+
return "audio/wav"
|
162 |
+
elif audio_format == "flac":
|
163 |
+
return "audio/flac"
|
164 |
+
elif audio_format == "mp3":
|
165 |
+
return "audio/mpeg"
|
166 |
+
else:
|
167 |
+
return "application/octet-stream"
|
168 |
+
|
169 |
+
|
170 |
+
@torch.inference_mode()
|
171 |
+
def inference(req: ServeTTSRequest):
|
172 |
+
|
173 |
+
idstr: str | None = req.reference_id
|
174 |
+
if idstr is not None:
|
175 |
+
ref_folder = Path("references") / idstr
|
176 |
+
ref_folder.mkdir(parents=True, exist_ok=True)
|
177 |
+
ref_audios = list_files(
|
178 |
+
ref_folder, AUDIO_EXTENSIONS, recursive=True, sort=False
|
179 |
+
)
|
180 |
+
prompt_tokens = [
|
181 |
+
encode_reference(
|
182 |
+
decoder_model=decoder_model,
|
183 |
+
reference_audio=audio_to_bytes(str(ref_audio)),
|
184 |
+
enable_reference_audio=True,
|
185 |
+
)
|
186 |
+
for ref_audio in ref_audios
|
187 |
+
]
|
188 |
+
prompt_texts = [
|
189 |
+
read_ref_text(str(ref_audio.with_suffix(".lab")))
|
190 |
+
for ref_audio in ref_audios
|
191 |
+
]
|
192 |
+
|
193 |
+
else:
|
194 |
+
# Parse reference audio aka prompt
|
195 |
+
refs = req.references
|
196 |
+
if refs is None:
|
197 |
+
refs = []
|
198 |
+
prompt_tokens = [
|
199 |
+
encode_reference(
|
200 |
+
decoder_model=decoder_model,
|
201 |
+
reference_audio=ref.audio,
|
202 |
+
enable_reference_audio=True,
|
203 |
+
)
|
204 |
+
for ref in refs
|
205 |
+
]
|
206 |
+
prompt_texts = [ref.text for ref in refs]
|
207 |
+
|
208 |
+
# LLAMA Inference
|
209 |
+
request = dict(
|
210 |
+
device=decoder_model.device,
|
211 |
+
max_new_tokens=req.max_new_tokens,
|
212 |
+
text=(
|
213 |
+
req.text
|
214 |
+
if not req.normalize
|
215 |
+
else ChnNormedText(raw_text=req.text).normalize()
|
216 |
+
),
|
217 |
+
top_p=req.top_p,
|
218 |
+
repetition_penalty=req.repetition_penalty,
|
219 |
+
temperature=req.temperature,
|
220 |
+
compile=args.compile,
|
221 |
+
iterative_prompt=req.chunk_length > 0,
|
222 |
+
chunk_length=req.chunk_length,
|
223 |
+
max_length=2048,
|
224 |
+
prompt_tokens=prompt_tokens,
|
225 |
+
prompt_text=prompt_texts,
|
226 |
+
)
|
227 |
+
|
228 |
+
response_queue = queue.Queue()
|
229 |
+
llama_queue.put(
|
230 |
+
GenerateRequest(
|
231 |
+
request=request,
|
232 |
+
response_queue=response_queue,
|
233 |
+
)
|
234 |
+
)
|
235 |
+
|
236 |
+
if req.streaming:
|
237 |
+
yield wav_chunk_header()
|
238 |
+
|
239 |
+
segments = []
|
240 |
+
while True:
|
241 |
+
result: WrappedGenerateResponse = response_queue.get()
|
242 |
+
if result.status == "error":
|
243 |
+
raise result.response
|
244 |
+
break
|
245 |
+
|
246 |
+
result: GenerateResponse = result.response
|
247 |
+
if result.action == "next":
|
248 |
+
break
|
249 |
+
|
250 |
+
with autocast_exclude_mps(
|
251 |
+
device_type=decoder_model.device.type, dtype=args.precision
|
252 |
+
):
|
253 |
+
fake_audios = decode_vq_tokens(
|
254 |
+
decoder_model=decoder_model,
|
255 |
+
codes=result.codes,
|
256 |
+
)
|
257 |
+
|
258 |
+
fake_audios = fake_audios.float().cpu().numpy()
|
259 |
+
|
260 |
+
if req.streaming:
|
261 |
+
yield (fake_audios * 32768).astype(np.int16).tobytes()
|
262 |
+
else:
|
263 |
+
segments.append(fake_audios)
|
264 |
+
|
265 |
+
if req.streaming:
|
266 |
+
return
|
267 |
+
|
268 |
+
if len(segments) == 0:
|
269 |
+
raise HTTPException(
|
270 |
+
HTTPStatus.INTERNAL_SERVER_ERROR,
|
271 |
+
content="No audio generated, please check the input text.",
|
272 |
+
)
|
273 |
+
|
274 |
+
fake_audios = np.concatenate(segments, axis=0)
|
275 |
+
yield fake_audios
|
276 |
+
|
277 |
+
|
278 |
+
async def inference_async(req: ServeTTSRequest):
|
279 |
+
for chunk in inference(req):
|
280 |
+
yield chunk
|
281 |
+
|
282 |
+
|
283 |
+
async def buffer_to_async_generator(buffer):
|
284 |
+
yield buffer
|
285 |
+
|
286 |
+
|
287 |
+
@routes.http.post("/v1/tts")
|
288 |
+
async def api_invoke_model(
|
289 |
+
req: Annotated[ServeTTSRequest, Body(exclusive=True)],
|
290 |
+
):
|
291 |
+
"""
|
292 |
+
Invoke model and generate audio
|
293 |
+
"""
|
294 |
+
|
295 |
+
if args.max_text_length > 0 and len(req.text) > args.max_text_length:
|
296 |
+
raise HTTPException(
|
297 |
+
HTTPStatus.BAD_REQUEST,
|
298 |
+
content=f"Text is too long, max length is {args.max_text_length}",
|
299 |
+
)
|
300 |
+
|
301 |
+
if req.streaming and req.format != "wav":
|
302 |
+
raise HTTPException(
|
303 |
+
HTTPStatus.BAD_REQUEST,
|
304 |
+
content="Streaming only supports WAV format",
|
305 |
+
)
|
306 |
+
|
307 |
+
if req.streaming:
|
308 |
+
return StreamResponse(
|
309 |
+
iterable=inference_async(req),
|
310 |
+
headers={
|
311 |
+
"Content-Disposition": f"attachment; filename=audio.{req.format}",
|
312 |
+
},
|
313 |
+
content_type=get_content_type(req.format),
|
314 |
+
)
|
315 |
+
else:
|
316 |
+
fake_audios = next(inference(req))
|
317 |
+
buffer = io.BytesIO()
|
318 |
+
sf.write(
|
319 |
+
buffer,
|
320 |
+
fake_audios,
|
321 |
+
decoder_model.spec_transform.sample_rate,
|
322 |
+
format=req.format,
|
323 |
+
)
|
324 |
+
|
325 |
+
return StreamResponse(
|
326 |
+
iterable=buffer_to_async_generator(buffer.getvalue()),
|
327 |
+
headers={
|
328 |
+
"Content-Disposition": f"attachment; filename=audio.{req.format}",
|
329 |
+
},
|
330 |
+
content_type=get_content_type(req.format),
|
331 |
+
)
|
332 |
+
|
333 |
+
|
334 |
+
@routes.http.post("/v1/health")
|
335 |
+
async def api_health():
|
336 |
+
"""
|
337 |
+
Health check
|
338 |
+
"""
|
339 |
+
|
340 |
+
return JSONResponse({"status": "ok"})
|
341 |
+
|
342 |
+
|
343 |
+
def parse_args():
|
344 |
+
parser = ArgumentParser()
|
345 |
+
parser.add_argument(
|
346 |
+
"--llama-checkpoint-path",
|
347 |
+
type=str,
|
348 |
+
default="checkpoints/fish-speech-1.4-sft-yth-lora",
|
349 |
+
)
|
350 |
+
parser.add_argument(
|
351 |
+
"--decoder-checkpoint-path",
|
352 |
+
type=str,
|
353 |
+
default="checkpoints/fish-speech-1.4/firefly-gan-vq-fsq-8x1024-21hz-generator.pth",
|
354 |
+
)
|
355 |
+
parser.add_argument("--decoder-config-name", type=str, default="firefly_gan_vq")
|
356 |
+
parser.add_argument("--device", type=str, default="cuda")
|
357 |
+
parser.add_argument("--half", action="store_true")
|
358 |
+
parser.add_argument("--compile", action="store_true")
|
359 |
+
parser.add_argument("--max-text-length", type=int, default=0)
|
360 |
+
parser.add_argument("--listen", type=str, default="127.0.0.1:8080")
|
361 |
+
parser.add_argument("--workers", type=int, default=1)
|
362 |
+
|
363 |
+
return parser.parse_args()
|
364 |
+
|
365 |
+
|
366 |
+
# Define Kui app
|
367 |
+
openapi = OpenAPI(
|
368 |
+
{
|
369 |
+
"title": "Fish Speech API",
|
370 |
+
},
|
371 |
+
).routes
|
372 |
+
|
373 |
+
|
374 |
+
class MsgPackRequest(HttpRequest):
|
375 |
+
async def data(self) -> Annotated[Any, ContentType("application/msgpack")]:
|
376 |
+
if self.content_type == "application/msgpack":
|
377 |
+
return ormsgpack.unpackb(await self.body)
|
378 |
+
|
379 |
+
raise HTTPException(
|
380 |
+
HTTPStatus.UNSUPPORTED_MEDIA_TYPE,
|
381 |
+
headers={"Accept": "application/msgpack"},
|
382 |
+
)
|
383 |
+
|
384 |
+
|
385 |
+
app = Kui(
|
386 |
+
routes=routes + openapi[1:], # Remove the default route
|
387 |
+
exception_handlers={
|
388 |
+
HTTPException: http_execption_handler,
|
389 |
+
Exception: other_exception_handler,
|
390 |
+
},
|
391 |
+
factory_class=FactoryClass(http=MsgPackRequest),
|
392 |
+
cors_config={},
|
393 |
+
)
|
394 |
+
|
395 |
+
|
396 |
+
if __name__ == "__main__":
|
397 |
+
|
398 |
+
import uvicorn
|
399 |
+
|
400 |
+
args = parse_args()
|
401 |
+
args.precision = torch.half if args.half else torch.bfloat16
|
402 |
+
|
403 |
+
logger.info("Loading Llama model...")
|
404 |
+
llama_queue = launch_thread_safe_queue(
|
405 |
+
checkpoint_path=args.llama_checkpoint_path,
|
406 |
+
device=args.device,
|
407 |
+
precision=args.precision,
|
408 |
+
compile=args.compile,
|
409 |
+
)
|
410 |
+
logger.info("Llama model loaded, loading VQ-GAN model...")
|
411 |
+
|
412 |
+
decoder_model = load_decoder_model(
|
413 |
+
config_name=args.decoder_config_name,
|
414 |
+
checkpoint_path=args.decoder_checkpoint_path,
|
415 |
+
device=args.device,
|
416 |
+
)
|
417 |
+
|
418 |
+
logger.info("VQ-GAN model loaded, warming up...")
|
419 |
+
|
420 |
+
# Dry run to check if the model is loaded correctly and avoid the first-time latency
|
421 |
+
list(
|
422 |
+
inference(
|
423 |
+
ServeTTSRequest(
|
424 |
+
text="Hello world.",
|
425 |
+
references=[],
|
426 |
+
reference_id=None,
|
427 |
+
max_new_tokens=1024,
|
428 |
+
chunk_length=200,
|
429 |
+
top_p=0.7,
|
430 |
+
repetition_penalty=1.2,
|
431 |
+
temperature=0.7,
|
432 |
+
emotion=None,
|
433 |
+
format="wav",
|
434 |
+
)
|
435 |
+
)
|
436 |
+
)
|
437 |
+
|
438 |
+
logger.info(f"Warming up done, starting server at http://{args.listen}")
|
439 |
+
host, port = args.listen.split(":")
|
440 |
+
uvicorn.run(app, host=host, port=int(port), workers=args.workers, log_level="info")
|