tp_inference device bug

shixian.shi

2023-03-09 c441eb08c44dfd4a7a8c68970fd3ebe7943d06ee

 funasr/bin/tp_inference.py

@@ -100,9 +100,9 @@
class SpeechText2Timestamp:
    def __init__(
        self,
        tp_train_config: Union[Path, str] = None,
        tp_model_file: Union[Path, str] = None,
        tp_cmvn_file: Union[Path, str] = None,
        timestamp_infer_config: Union[Path, str] = None,
        timestamp_model_file: Union[Path, str] = None,
        timestamp_cmvn_file: Union[Path, str] = None,
        device: str = "cpu",
        dtype: str = "float32",
        **kwargs,
@@ -110,11 +110,14 @@
        assert check_argument_types()
        # 1. Build ASR model
        tp_model, tp_train_args = ASRTask.build_model_from_file(
            tp_train_config, tp_model_file, device
            timestamp_infer_config, timestamp_model_file, device
        )
        if 'cuda' in device:
            tp_model = tp_model.cuda()
            
        frontend = None
        if tp_train_args.frontend is not None:
            frontend = WavFrontend(cmvn_file=tp_cmvn_file, **tp_train_args.frontend_conf)
            frontend = WavFrontend(cmvn_file=timestamp_cmvn_file, **tp_train_args.frontend_conf)
        
        logging.info("tp_model: {}".format(tp_model))
        logging.info("tp_train_args: {}".format(tp_train_args))
@@ -148,11 +151,11 @@
        # Input as audio signal
        if isinstance(speech, np.ndarray):
            speech = torch.tensor(speech)

        if self.frontend is not None:
            feats, feats_len = self.frontend.forward(speech, speech_lengths)
            feats = to_device(feats, device=self.device)
            feats_len = feats_len.int()
            self.tp_model.frontend = None
        else:
            feats = speech
            feats_len = speech_lengths
@@ -178,9 +181,9 @@
        ngpu: int,
        log_level: Union[int, str],
        data_path_and_name_and_type,
        tp_train_config: Optional[str],
        tp_model_file: Optional[str],
        tp_cmvn_file: Optional[str] = None,
        timestamp_infer_config: Optional[str],
        timestamp_model_file: Optional[str],
        timestamp_cmvn_file: Optional[str] = None,
        raw_inputs: Union[np.ndarray, torch.Tensor] = None,
        key_file: Optional[str] = None,
        allow_variable_data_keys: bool = False,
@@ -194,9 +197,9 @@
        batch_size=batch_size,
        ngpu=ngpu,
        log_level=log_level,
        tp_train_config=tp_train_config,
        tp_model_file=tp_model_file,
        tp_cmvn_file=tp_cmvn_file,
        timestamp_infer_config=timestamp_infer_config,
        timestamp_model_file=timestamp_model_file,
        timestamp_cmvn_file=timestamp_cmvn_file,
        key_file=key_file,
        allow_variable_data_keys=allow_variable_data_keys,
        output_dir=output_dir,
@@ -213,9 +216,9 @@
        ngpu: int,
        log_level: Union[int, str],
        # data_path_and_name_and_type,
        tp_train_config: Optional[str],
        tp_model_file: Optional[str],
        tp_cmvn_file: Optional[str] = None,
        timestamp_infer_config: Optional[str],
        timestamp_model_file: Optional[str],
        timestamp_cmvn_file: Optional[str] = None,
        # raw_inputs: Union[np.ndarray, torch.Tensor] = None,
        key_file: Optional[str] = None,
        allow_variable_data_keys: bool = False,
@@ -240,15 +243,14 @@
        device = "cuda"
    else:
        device = "cpu"

    # 1. Set random-seed
    set_all_random_seed(seed)

    # 2. Build speech2vadsegment
    speechtext2timestamp_kwargs = dict(
        tp_train_config=tp_train_config,
        tp_model_file=tp_model_file,
        tp_cmvn_file=tp_cmvn_file,
        timestamp_infer_config=timestamp_infer_config,
        timestamp_model_file=timestamp_model_file,
        timestamp_cmvn_file=timestamp_cmvn_file,
        device=device,
        dtype=dtype,
    )
@@ -365,17 +367,17 @@

    group = parser.add_argument_group("The model configuration related")
    group.add_argument(
        "--tp_train_config",
        "--timestamp_infer_config",
        type=str,
        help="VAD infer configuration",
    )
    group.add_argument(
        "--tp_model_file",
        "--timestamp_model_file",
        type=str,
        help="VAD model parameter file",
    )
    group.add_argument(
        "--tp_cmvn_file",
        "--timestamp_cmvn_file",
        type=str,
        help="Global cmvn file",
    )