vad conf

游雁

2024-03-19 d29f201e3201bde6a984e436888a2aae877e449f

vad conf

 examples/industrial_data_pretraining/emotion2vec/demo.py

@@ -9,7 +9,7 @@
model = AutoModel(model="iic/emotion2vec_base_finetuned", model_revision="v2.0.4",
                  # vad_model="iic/speech_fsmn_vad_zh-cn-16k-common-pytorch",
                  # vad_model_revision="v2.0.4",
                  # vad_kwargs={"max_single_segment_time": 10},
                  # vad_kwargs={"max_single_segment_time": 1000},
                  )

wav_file = f"{model.model_path}/example/test.wav"

 examples/industrial_data_pretraining/paraformer/demo.py

@@ -9,7 +9,7 @@
                  model_revision="v2.0.4",
                  vad_model="iic/speech_fsmn_vad_zh-cn-16k-common-pytorch",
                  vad_model_revision="v2.0.4",
                  vad_kwargs={"max_single_segment_time": 60},
                  vad_kwargs={"max_single_segment_time": 60000},
                  punc_model="iic/punc_ct-transformer_zh-cn-common-vocab272727-pytorch",
                  punc_model_revision="v2.0.4",
                  # spk_model="iic/speech_campplus_sv_zh-cn_16k-common",

 examples/industrial_data_pretraining/whisper/demo.py

@@ -10,7 +10,7 @@
model = AutoModel(model="iic/Whisper-large-v3",
                  model_revision="v2.0.5",
                  vad_model="iic/speech_fsmn_vad_zh-cn-16k-common-pytorch",
                  vad_kwargs={"max_single_segment_time": 30},
                  vad_kwargs={"max_single_segment_time": 30000},
                  )

res = model.generate(

 examples/industrial_data_pretraining/whisper/demo_from_openai.py

@@ -12,7 +12,7 @@
# model = AutoModel(model="Whisper-large-v2", hub="openai")
model = AutoModel(model="Whisper-large-v3",
                  vad_model="iic/speech_fsmn_vad_zh-cn-16k-common-pytorch",
                  vad_kwargs={"max_single_segment_time": 30},
                  vad_kwargs={"max_single_segment_time": 30000},
                  hub="openai",
                  )


 funasr/auto/auto_model.py

@@ -312,7 +312,8 @@
            key = res[i]["key"]
            vadsegments = res[i]["value"]
            input_i = data_list[i]
            speech = load_audio_text_image_video(input_i, fs=kwargs["frontend"].fs, audio_fs=kwargs.get("fs", 16000))
            fs = kwargs["frontend"].fs if hasattr(kwargs["frontend"], "fs") else 16000
            speech = load_audio_text_image_video(input_i, fs=fs, audio_fs=kwargs.get("fs", 16000))
            speech_lengths = len(speech)
            n = len(vadsegments)
            data_with_index = [(vadsegments[i], i) for i in range(n)]