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
atsr
语帆6d7b94571032
diff --git a/funasr/auto/auto_model.py b/funasr/auto/auto_model.pyindex 56dd5b5e4..9bb9ce07d 100644--- a/funasr/auto/auto_model.py+++ b/funasr/auto/auto_model.py@@ -41,7 +41,7 @@ def prepare_data_iterator(data_in, input_len=None, data_type=None, key=None):chars = string.ascii_letters + string.digitsif isinstance(data_in, str) and data_in.startswith('http'): # urldata_in = download_from_url(data_in)- pdb.set_trace()+if isinstance(data_in, str) and os.path.exists(data_in): # wav_path; filelist: wav.scp, file.jsonl;text.txt;_, file_extension = os.path.splitext(data_in)file_extension = file_extension.lower()diff --git a/funasr/models/lcbnet/model.py b/funasr/models/lcbnet/model.pyindex ab557e6d8..09e6dd137 100644--- a/funasr/models/lcbnet/model.py+++ b/funasr/models/lcbnet/model.py@@ -426,6 +426,7 @@ class LCBNet(nn.Module):tokenizer=tokenizer)time2 = time.perf_counter()meta_data["load_data"] = f"{time2 - time1:0.3f}"+ pdb.set_trace()audio_sample_list = sample_list[0]ocr_sample_list = sample_list[1]speech, speech_lengths = extract_fbank(audio_sample_list, data_type=kwargs.get("data_type", "sound"),@@ -441,7 +442,7 @@ class LCBNet(nn.Module):encoder_out, encoder_out_lens = self.encode(speech, speech_lengths)if isinstance(encoder_out, tuple):encoder_out = encoder_out[0]-+ pdb.set_trace()ocr_list_new = [[x + 1 if x != 0 else x for x in sublist] for sublist in ocr_sample_list]ocr = torch.tensor(ocr_list_new).to(device=kwargs["device"])ocr_lengths = ocr.new_full([1], dtype=torch.long, fill_value=ocr.size(1)).to(device=kwargs["device"])diff --git a/funasr/utils/load_utils.py b/funasr/utils/load_utils.pyindex ccb5670c2..644af2324 100644--- a/funasr/utils/load_utils.py+++ b/funasr/utils/load_utils.py@@ -31,7 +31,7 @@ def load_audio_text_image_video(data_or_path_or_list, fs: int = 16000, audio_fs:return [load_audio_text_image_video(audio, fs=fs, audio_fs=audio_fs, data_type=data_type, **kwargs) for audio in data_or_path_or_list]if isinstance(data_or_path_or_list, str) and data_or_path_or_list.startswith('http'): # download url to local filedata_or_path_or_list = download_from_url(data_or_path_or_list)- pdb.set_trace()+if isinstance(data_or_path_or_list, str) and os.path.exists(data_or_path_or_list): # local fileif data_type is None or data_type == "sound":data_or_path_or_list, audio_fs = torchaudio.load(data_or_path_or_list)@@ -67,7 +67,7 @@ def load_audio_text_image_video(data_or_path_or_list, fs: int = 16000, audio_fs:else:pass# print(f"unsupport data type: {data_or_path_or_list}, return raw data")- pdb.set_trace()+if audio_fs != fs and data_type != "text":resampler = torchaudio.transforms.Resample(audio_fs, fs)data_or_path_or_list = resampler(data_or_path_or_list[None, :])[0, :]