funasr2

游雁

2023-12-19 0e622e694e6cb4459955f1e5942a7c53349ce640

 funasr/datasets/audio_datasets/load_audio_extract_fbank.py

File was renamed from funasr/datasets/fun_datasets/load_audio_extract_fbank.py
@@ -46,15 +46,16 @@
   array = np.frombuffer((middle_data.astype(dtype) - offset) / abs_max, dtype=np.float32)
   return array

def extract_fbank(data, data_len = None, date_type: str="sound", frontend=None):
	
def extract_fbank(data, data_len = None, data_type: str="sound", frontend=None):
   # import pdb;
   # pdb.set_trace()
   if isinstance(data, np.ndarray):
      data = torch.from_numpy(data)
      if len(data) < 2:
      if len(data.shape) < 2:
         data = data[None, :] # data: [batch, N]
      data_len = [data.shape[1]] if data_len is None else data_len
   elif isinstance(data, torch.Tensor):
      if len(data) < 2:
      if len(data.shape) < 2:
         data = data[None, :] # data: [batch, N]
      data_len = [data.shape[1]] if data_len is None else data_len
   elif isinstance(data, (list, tuple)):
@@ -67,7 +68,7 @@
      data = pad_sequence(data_list, batch_first=True) # data: [batch, N]
   # import pdb;
   # pdb.set_trace()
   if date_type == "sound":
   if data_type == "sound":
      data, data_len = frontend(data, data_len)
   
   if isinstance(data_len, (list, tuple)):