FunASR

FunASR copied to clipboard

实时语音识别示例代码报错

Before asking:

1. search the issues.
2. search the docs.

Code

from funasr import AutoModel
from funasr.utils.postprocess_utils import rich_transcription_postprocess

model_dir = "iic/SenseVoiceSmall"

chunk_size = [0, 10, 5] #[0, 10, 5] 600ms, [0, 8, 4] 480ms
encoder_chunk_look_back = 4 #number of chunks to lookback for encoder self-attention
decoder_chunk_look_back = 1 #number of encoder chunks to lookback for decoder cross-attention

model = AutoModel(model="paraformer-zh-streaming",model_revision="v2.0.4",disable_update=True)

import soundfile
import os

#wav_file = os.path.join(model.model_path, "example/asr_example.wav")
wav_file = "./test_vad.wav"
speech, sample_rate = soundfile.read(wav_file)
chunk_stride = chunk_size[1] * 960 # 600ms
cache = {}
total_chunk_num = int(len((speech)-1)/chunk_stride+1)
print(f"total_chunk_num={total_chunk_num},chunk_stride={chunk_stride}")

for i in range(total_chunk_num):
    speech_chunk = speech[i*chunk_stride:(i+1)*chunk_stride]
    is_final = i == total_chunk_num - 1
    print(f"i={i},is_final={is_final},speech_chunk={len(speech_chunk)}")
    res = model.generate(input=speech_chunk, 
                         cache=cache, 
                         is_final=is_final, 
                         chunk_size=chunk_size, 
                         encoder_chunk_look_back=encoder_chunk_look_back, 
                         decoder_chunk_look_back=decoder_chunk_look_back)
    print(res)

What's your environment?

• OS (e.g., Linux): Windows11
• FunASR Version (e.g., 1.0.0): 1.2.4
• ModelScope Version (e.g., 1.11.0): 1.23.1
• PyTorch Version (e.g., 2.0.0): 2.6.0
• How you installed funasr (pip, source): pip
• Python version: 3.10.16
• GPU (e.g., V100M32)
• CUDA/cuDNN version (e.g., cuda11.7):
• Docker version (e.g., funasr-runtime-sdk-cpu-0.4.1)
• Any other relevant information:

报错信息

ValueError                                Traceback (most recent call last)
Cell In[38], line 5
      3 is_final = i == total_chunk_num - 1
      4 print(f"i={i},is_final={is_final},speech_chunk={len(speech_chunk)}")
----> 5 res = model.generate(input=speech_chunk, 
      6                      cache=cache, 
      7                      is_final=is_final, 
      8                      chunk_size=chunk_size, 
      9                      encoder_chunk_look_back=encoder_chunk_look_back, 
     10                      decoder_chunk_look_back=decoder_chunk_look_back)
     11 print(res)
File c:\xxx\lib\site-packages\funasr\auto\auto_model.py:303, in AutoModel.generate(self, input, input_len, **cfg)
    301 def generate(self, input, input_len=None, **cfg):
    302     if self.vad_model is None:
--> 303         return self.inference(input, input_len=input_len, **cfg)
    305     else:
    306         return self.inference_with_vad(input, input_len=input_len, **cfg)
File c:\xxx\lib\site-packages\funasr\auto\auto_model.py:345, in AutoModel.inference(self, input, input_len, model, kwargs, key, **cfg)
    343 time1 = time.perf_counter()
    344 with torch.no_grad():
--> 345     res = model.inference(**batch, **kwargs)
    346     if isinstance(res, (list, tuple)):
...
--> 437     batch_size, timesteps, input_dim = x.size()
    438     start_idx = 0
    439     if cache is not None: