Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.
Python asraudiochineseemotion-recognitionfunasrmcp-servermultilingual-asropenai-compatible-apiparaformerpunctuationpytorchreal-time-asrspeaker-diarizationspeech-recognitionspeech-to-textstreaming-asrtranscriptionvllmvoice-activity-detectionwhisper-alternative
fix(load): clear FileNotFoundError for missing audio path (#2981)
zhifu gaoe6ab4c37040e
diff --git a/funasr/utils/load_utils.py b/funasr/utils/load_utils.pyindex 03acf0df4..e30559a2a 100644--- a/funasr/utils/load_utils.py+++ b/funasr/utils/load_utils.py@@ -97,6 +97,19 @@ def load_audio_text_image_video(): # download url to local filedata_or_path_or_list = download_from_url(data_or_path_or_list)+ # Fail fast with a clear error if an audio file path does not exist, instead of+ # silently passing the string downstream (which later crashes with a cryptic+ # "expected Tensor ... but got str" deep inside the model).+ if (+ isinstance(data_or_path_or_list, str)+ and data_type in (None, "sound")+ and not data_or_path_or_list.startswith(("http://", "https://"))+ and not os.path.exists(data_or_path_or_list)+ ):+ raise FileNotFoundError(+ f"Audio file not found: {data_or_path_or_list!r}. Pass a valid local file "+ f"path, URL, numpy array, torch.Tensor, or bytes."+ )if (isinstance(data_or_path_or_list, str) and os.path.exists(data_or_path_or_list)) or hasattr(data_or_path_or_list, 'read'): # local file or bytes ioif data_type is None or data_type == "sound":if hasattr(data_or_path_or_list, "read") and hasattr(data_or_path_or_list, "seek"):