Set default whisper device to 'auto' in corpus_builder.py

This commit is contained in:
2026-03-19 23:37:24 +01:00
parent e58b466b2c
commit 7c02044f4d

View File

@@ -1534,7 +1534,7 @@ def process_media(path: Path, args) -> List[Record]:
slices = slice_audio(wav_path, slice_dir, nslices, args.overlap_sec, ffprobe_bin, ffmpeg_bin)
mpw = args.mp_workers or len(slices)
device = _resolve_whisper_device(args.whisper_device)
device = args.whisper_device or "auto"
ctx = mp.get_context("spawn")
pool = ctx.Pool(processes=mpw, initializer=_whisper_pool_init, initargs=(args.whisper_model, device))
try: