speech_recognition/FunASR
1
0
Fork 0
mirror of https://github.com/modelscope/FunASR synced 2025-09-15 14:48:36 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

update vad inference

Browse Source
This commit is contained in:
凌匀 2023-02-28 14:33:40 +08:00
parent 2ed32a5d40
commit 3388361d3b
2 changed files with 7 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
funasr/bin/vad_inference.py
View File

@ -86,7 +86,8 @@ class Speech2VadSegment:
@torch.no_grad()
def __call__(
self, speech: Union[torch.Tensor, np.ndarray], speech_lengths: Union[torch.Tensor, np.ndarray] = None
self, speech: Union[torch.Tensor, np.ndarray], speech_lengths: Union[torch.Tensor, np.ndarray] = None,
in_cache: Dict[str, torch.Tensor] = dict()
) -> Tuple[List[List[int]], Dict[str, torch.Tensor]]:
"""Inference
@ -125,11 +126,12 @@ class Speech2VadSegment:
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": is_final
"is_final": is_final,
"in_cache": in_cache
}
# a. To device
batch = to_device(batch, device=self.device)
segments_part = self.vad_model(**batch)
segments_part, in_cache = self.vad_model(**batch)
if segments_part:
for batch_num in range(0, self.batch_size):
segments[batch_num] += segments_part[batch_num]

5
funasr/models/e2e_vad.py
View File

@ -444,7 +444,7 @@ class E2EVadModel(nn.Module):
def forward(self, feats: torch.Tensor, waveform: torch.tensor, in_cache: Dict[str, torch.Tensor] = dict(),
is_final: bool = False
) -> List[List[List[int]]]:
) -> Tuple[List[List[List[int]]], Dict[str, torch.Tensor]]:
self.waveform = waveform # compute decibel for each frame
self.ComputeDecibel()
self.ComputeScores(feats, in_cache)
@ -468,8 +468,7 @@ class E2EVadModel(nn.Module):
if is_final:
# reset class variables and clear the dict for the next query
self.AllResetDetection()
in_cache.clear()
return segments
return segments, in_cache
def DetectCommonFrames(self) -> int:
if self.vad_state_machine == VadStateMachine.kVadInStateEndPointDetected: