From 2cdb2d654f2109ef4e648bae6f169143e267e5db Mon Sep 17 00:00:00 2001
From: zhuzizyf <42790740+zhuzizyf@users.noreply.github.com>
Date: 星期六, 11 三月 2023 14:33:14 +0800
Subject: [PATCH] Update dataset.py
---
funasr/bin/asr_inference_paraformer_vad_punc.py | 96 -----------------------------------------------
1 files changed, 1 insertions(+), 95 deletions(-)
diff --git a/funasr/bin/asr_inference_paraformer_vad_punc.py b/funasr/bin/asr_inference_paraformer_vad_punc.py
index 96f70ef..1320877 100644
--- a/funasr/bin/asr_inference_paraformer_vad_punc.py
+++ b/funasr/bin/asr_inference_paraformer_vad_punc.py
@@ -43,6 +43,7 @@
from funasr.utils import asr_utils, wav_utils, postprocess_utils
from funasr.models.frontend.wav_frontend import WavFrontend
from funasr.tasks.vad import VADTask
+from funasr.bin.vad_inference import Speech2VadSegment
from funasr.utils.timestamp_tools import time_stamp_lfr6_pl
from funasr.bin.punctuation_infer import Text2Punc
from funasr.models.e2e_asr_paraformer import BiCifParaformer, ContextualParaformer
@@ -363,101 +364,6 @@
else:
hotword_list = None
return hotword_list
-
-class Speech2VadSegment:
- """Speech2VadSegment class
-
- Examples:
- >>> import soundfile
- >>> speech2segment = Speech2VadSegment("vad_config.yml", "vad.pt")
- >>> audio, rate = soundfile.read("speech.wav")
- >>> speech2segment(audio)
- [[10, 230], [245, 450], ...]
-
- """
-
- def __init__(
- self,
- vad_infer_config: Union[Path, str] = None,
- vad_model_file: Union[Path, str] = None,
- vad_cmvn_file: Union[Path, str] = None,
- device: str = "cpu",
- batch_size: int = 1,
- dtype: str = "float32",
- **kwargs,
- ):
- assert check_argument_types()
-
- # 1. Build vad model
- vad_model, vad_infer_args = VADTask.build_model_from_file(
- vad_infer_config, vad_model_file, device
- )
- frontend = None
- if vad_infer_args.frontend is not None:
- frontend = WavFrontend(cmvn_file=vad_cmvn_file, **vad_infer_args.frontend_conf)
-
- # logging.info("vad_model: {}".format(vad_model))
- # logging.info("vad_infer_args: {}".format(vad_infer_args))
- vad_model.to(dtype=getattr(torch, dtype)).eval()
-
- self.vad_model = vad_model
- self.vad_infer_args = vad_infer_args
- self.device = device
- self.dtype = dtype
- self.frontend = frontend
- self.batch_size = batch_size
-
- @torch.no_grad()
- def __call__(
- self, speech: Union[torch.Tensor, np.ndarray], speech_lengths: Union[torch.Tensor, np.ndarray] = None
- ) -> List[List[int]]:
- """Inference
-
- Args:
- speech: Input speech data
- Returns:
- text, token, token_int, hyp
-
- """
- assert check_argument_types()
-
- # Input as audio signal
- if isinstance(speech, np.ndarray):
- speech = torch.tensor(speech)
-
- if self.frontend is not None:
- self.frontend.filter_length_max = math.inf
- fbanks, fbanks_len = self.frontend.forward_fbank(speech, speech_lengths)
- feats, feats_len = self.frontend.forward_lfr_cmvn(fbanks, fbanks_len)
- fbanks = to_device(fbanks, device=self.device)
- feats = to_device(feats, device=self.device)
- feats_len = feats_len.int()
- else:
- raise Exception("Need to extract feats first, please configure frontend configuration")
-
- # b. Forward Encoder streaming
- t_offset = 0
- step = min(feats_len, 6000)
- segments = [[]] * self.batch_size
- for t_offset in range(0, feats_len, min(step, feats_len - t_offset)):
- if t_offset + step >= feats_len - 1:
- step = feats_len - t_offset
- is_final_send = True
- else:
- is_final_send = False
- batch = {
- "feats": feats[:, t_offset:t_offset + step, :],
- "waveform": speech[:, t_offset * 160:min(speech.shape[-1], (t_offset + step - 1) * 160 + 400)],
- "is_final_send": is_final_send
- }
- # a. To device
- batch = to_device(batch, device=self.device)
- segments_part = self.vad_model(**batch)
- if segments_part:
- for batch_num in range(0, self.batch_size):
- segments[batch_num] += segments_part[batch_num]
-
- return fbanks, segments
def inference(
--
Gitblit v1.9.1