scott
29b66e24bb
All checks were successful
Build ROCm Image / build (push) Successful in 3m17s
Voice conditionals (s3tokenizer + voice encoder + mel embeddings) are expensive to compute but depend only on the reference audio, not the text. Previously they ran on every synthesis chunk — 3x wasted work for a 3-chunk request. Now computed once at startup and reused. Also wrap generate() in torch.amp.autocast(float16) for ~2x speedup on all model computation (T3 LLM, S3Gen CFM, HiFiGAN vocoder). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
67 lines
2.0 KiB
Python
67 lines
2.0 KiB
Python
import asyncio
|
|
import logging
|
|
import sys
|
|
from functools import partial
|
|
|
|
from wyoming.server import AsyncServer
|
|
|
|
import engine
|
|
from config import get_wyoming_host, get_wyoming_port, load_config
|
|
from wyoming_handler import ChatterboxWyomingHandler
|
|
from wyoming_voices import create_wyoming_info, load_voices
|
|
|
|
logging.basicConfig(
|
|
level=logging.INFO,
|
|
format="%(asctime)s %(levelname)s %(name)s: %(message)s",
|
|
stream=sys.stdout,
|
|
)
|
|
logger = logging.getLogger(__name__)
|
|
|
|
|
|
def _warmup(voices: dict) -> None:
|
|
"""Pre-compute voice conditionals and populate MIOpen's kernel cache."""
|
|
from wyoming_voices import resolve_voice
|
|
|
|
# Pre-compute conditionals for all discovered voices so the first real
|
|
# request doesn't pay the s3tokenizer + voice encoder cost.
|
|
for name, path in voices.items():
|
|
try:
|
|
engine.prepare_voice(path)
|
|
except Exception:
|
|
logger.warning(f"Failed to prepare voice '{name}' (non-fatal)", exc_info=True)
|
|
|
|
# Synthesis warmup to populate MIOpen's in-memory kernel cache.
|
|
audio_prompt = resolve_voice(None, voices) if voices else None
|
|
logger.info("Running warmup synthesis...")
|
|
try:
|
|
engine.synthesize(text="Warmup.", audio_prompt_path=audio_prompt)
|
|
logger.info("Warmup complete")
|
|
except Exception:
|
|
logger.warning("Warmup synthesis failed (non-fatal)", exc_info=True)
|
|
|
|
|
|
async def main() -> None:
|
|
load_config()
|
|
|
|
logger.info("Loading TTS model...")
|
|
if not engine.load_model():
|
|
logger.error("Failed to load model, exiting")
|
|
sys.exit(1)
|
|
|
|
voices = load_voices()
|
|
wyoming_info = create_wyoming_info(engine.get_sample_rate(), voices)
|
|
|
|
_warmup(voices)
|
|
|
|
host = get_wyoming_host()
|
|
port = get_wyoming_port()
|
|
uri = f"tcp://{host}:{port}"
|
|
|
|
logger.info(f"Starting Wyoming server on {uri}")
|
|
server = AsyncServer.from_uri(uri)
|
|
await server.run(partial(ChatterboxWyomingHandler, wyoming_info, voices))
|
|
|
|
|
|
if __name__ == "__main__":
|
|
asyncio.run(main())
|