| | |
|---|
| | | for audio in data_or_path_or_list |
|---|
| | | ] |
|---|
| | | if isinstance(data_or_path_or_list, str) and data_or_path_or_list.startswith( |
|---|
| | | ("http://", "https://") |
|---|
| | | ("http://", "https://") |
|---|
| | | ): # download url to local file |
|---|
| | | data_or_path_or_list = download_from_url(data_or_path_or_list) |
|---|
| | | |
|---|
| | |
|---|
| | | elif isinstance(data_or_path_or_list, str) and data_type == "text" and tokenizer is not None: |
|---|
| | | data_or_path_or_list = tokenizer.encode(data_or_path_or_list) |
|---|
| | | elif isinstance(data_or_path_or_list, np.ndarray): # audio sample point |
|---|
| | | data_or_path_or_list = torch.from_numpy(data_or_path_or_list).squeeze() # [n_samples,] |
|---|
| | | data_or_path_or_list = torch.from_numpy(data_or_path_or_list) # .squeeze() # [n_samples,] |
|---|
| | | elif isinstance(data_or_path_or_list, str) and data_type == "kaldi_ark": |
|---|
| | | data_mat = kaldiio.load_mat(data_or_path_or_list) |
|---|
| | | if isinstance(data_mat, tuple): |
|---|
