wx44wx/pipecat
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Smart Turn v3.1

Browse Source
This commit is contained in:
marcus-daily
2025-12-03 18:59:05 +00:00
committed by Marcus
parent b5e79f9dc5
commit 39bca074d7
3 changed files with 6 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
pyproject.toml
View File

@@ -159,7 +159,7 @@ where = ["src"]
"src/pipecat/audio/dtmf/dtmf-star.wav", "src/pipecat/audio/dtmf/dtmf-star.wav",
] ]
"pipecat.services.aws_nova_sonic" = ["src/pipecat/services/aws_nova_sonic/ready.wav"] "pipecat.services.aws_nova_sonic" = ["src/pipecat/services/aws_nova_sonic/ready.wav"]
"pipecat.audio.turn.smart_turn.data" = ["src/pipecat/audio/turn/smart_turn/data/smart-turn-v3.0.onnx"] "pipecat.audio.turn.smart_turn.data" = ["src/pipecat/audio/turn/smart_turn/data/smart-turn-v3.1-cpu.onnx"]
[tool.pytest.ini_options] [tool.pytest.ini_options]
addopts = "--verbose" addopts = "--verbose"

BIN
src/pipecat/audio/turn/smart_turn/data/smart-turn-v3.0.onnx → src/pipecat/audio/turn/smart_turn/data/smart-turn-v3.1-cpu.onnx
View File

Binary file not shown.

10
src/pipecat/audio/turn/smart_turn/local_smart_turn_v3.py
View File

@@ -42,17 +42,15 @@ class LocalSmartTurnAnalyzerV3(BaseSmartTurn):
Args: Args:
smart_turn_model_path: Path to the ONNX model file. If this is not smart_turn_model_path: Path to the ONNX model file. If this is not
set, the bundled smart-turn-v3.0 model will be used. set, the bundled smart-turn-v3.1-cpu model will be used.
cpu_count: The number of CPUs to use for inference. Defaults to 1. cpu_count: The number of CPUs to use for inference. Defaults to 1.
**kwargs: Additional arguments passed to BaseSmartTurn. **kwargs: Additional arguments passed to BaseSmartTurn.
""" """
super().__init__(**kwargs) super().__init__(**kwargs)
logger.debug("Loading Local Smart Turn v3 model...")
if not smart_turn_model_path: if not smart_turn_model_path:
# Load bundled model # Load bundled model
model_name = "smart-turn-v3.0.onnx" model_name = "smart-turn-v3.1-cpu.onnx"
package_path = "pipecat.audio.turn.smart_turn.data" package_path = "pipecat.audio.turn.smart_turn.data"
try: try:
@@ -70,6 +68,8 @@ class LocalSmartTurnAnalyzerV3(BaseSmartTurn):
impresources.files(package_path).joinpath(model_name) impresources.files(package_path).joinpath(model_name)
) )
logger.debug(f"Loading Local Smart Turn v3.x model from {smart_turn_model_path}...")
so = ort.SessionOptions() so = ort.SessionOptions()
so.execution_mode = ort.ExecutionMode.ORT_SEQUENTIAL so.execution_mode = ort.ExecutionMode.ORT_SEQUENTIAL
so.inter_op_num_threads = 1 so.inter_op_num_threads = 1
@@ -79,7 +79,7 @@ class LocalSmartTurnAnalyzerV3(BaseSmartTurn):
self._feature_extractor = WhisperFeatureExtractor(chunk_length=8) self._feature_extractor = WhisperFeatureExtractor(chunk_length=8)
self._session = ort.InferenceSession(smart_turn_model_path, sess_options=so) self._session = ort.InferenceSession(smart_turn_model_path, sess_options=so)
logger.debug("Loaded Local Smart Turn v3") logger.debug("Loaded Local Smart Turn v3.x")
def _predict_endpoint(self, audio_array: np.ndarray) -> Dict[str, Any]: def _predict_endpoint(self, audio_array: np.ndarray) -> Dict[str, Any]:
"""Predict end-of-turn using local ONNX model.""" """Predict end-of-turn using local ONNX model."""