fix(onnx-asr): исправлены баги интеграции после smoke-теста

- Зачем:
  - при smoke-тесте выявлены расхождения между onnx-asr API и мок-тестами.
- Что:
  - убран deprecated cpu_preprocessing из load_model.
  - sample_rate передан как keyword-аргумент в recognize().
  - исправлены имена полей SegmentResult: start_ts→start, end_ts→end.
  - добавлен device "onnx" в DEVICE_DEFAULTS и _VALID_DEVICES.
- Проверка:
  - uv run pytest -q (172 passed).
  - uv run transcribe file.mp4 --device onnx --model gigaam-v3.
This commit is contained in:
2026-04-25 21:48:54 +03:00
parent 7d39ab908b
commit b1cbdc3e0b
3 changed files with 9 additions and 11 deletions
+3 -4
View File
@@ -59,7 +59,6 @@ class OnnxAsrBackend:
model = onnx_asr.load_model(
model=model_path,
quantization=ct,
cpu_preprocessing=True,
)
vad = onnx_asr.load_vad("silero")
self._vad = vad
@@ -89,10 +88,10 @@ class OnnxAsrBackend:
segments: list[Segment] = []
detected_language = language or "unknown"
for vad_seg in model.recognize(audio_array, 16000, language=language):
for vad_seg in model.recognize(audio_array, sample_rate=16000, language=language):
seg = Segment(
start=max(0.0, vad_seg.start_ts),
end=max(0.0, vad_seg.end_ts),
start=max(0.0, vad_seg.start),
end=max(0.0, vad_seg.end),
text=vad_seg.text,
)
if on_segment is not None:
+2 -1
View File
@@ -22,11 +22,12 @@ DEVICE_DEFAULTS: dict[str, dict[str, str]] = {
"openvino": {"model": "medium", "compute_type": "int8"},
"openvino-gpu": {"model": "medium", "compute_type": "int8"},
"openvino-cpu": {"model": "medium", "compute_type": "int8"},
"onnx": {"model": "gigaam-v3", "compute_type": "int8"},
}
# Одно место правды для допустимых ключей конфига
_VALID_KEYS = set(HARDCODED_DEFAULTS)
_VALID_DEVICES = {"auto", "cpu", "cuda", "openvino", "openvino-gpu", "openvino-cpu"}
_VALID_DEVICES = {"auto", "cpu", "cuda", "openvino", "openvino-gpu", "openvino-cpu", "onnx"}
def find_config_file() -> Path | None: