support wav_file input

haoneng.lhn

2023-04-14 5589b4a6177a0ca6836c304a152befe87f0ddd96

 funasr/bin/asr_inference_paraformer_streaming.py

@@ -19,6 +19,7 @@

import numpy as np
import torch
import torchaudio
from typeguard import check_argument_types

from funasr.fileio.datadir_writer import DatadirWriter
@@ -607,17 +608,21 @@
    ):

        # 3. Build data-iterator
        if data_path_and_name_and_type is not None and data_path_and_name_and_type[2] == "bytes":
            raw_inputs = _load_bytes(data_path_and_name_and_type[0])
            raw_inputs = torch.tensor(raw_inputs)
        if data_path_and_name_and_type is None and raw_inputs is not None:
            if isinstance(raw_inputs, np.ndarray):
                raw_inputs = torch.tensor(raw_inputs)
        is_final = False
        if param_dict is not None and "cache" in param_dict:
            cache = param_dict["cache"]
        if param_dict is not None and "is_final" in param_dict:
            is_final = param_dict["is_final"]

        if data_path_and_name_and_type is not None and data_path_and_name_and_type[2] == "bytes":
            raw_inputs = _load_bytes(data_path_and_name_and_type[0])
            raw_inputs = torch.tensor(raw_inputs)
        if data_path_and_name_and_type is not None and data_path_and_name_and_type[2] == "sound":
            raw_inputs = torchaudio.load(data_path_and_name_and_type[0])[0][0]
            is_final = True
        if data_path_and_name_and_type is None and raw_inputs is not None:
            if isinstance(raw_inputs, np.ndarray):
                raw_inputs = torch.tensor(raw_inputs)
        # 7 .Start for-loop
        # FIXME(kamo): The output format should be discussed about
        asr_result_list = []