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

Merge branch 'dev_cmz2' of github.com:alibaba-damo-academy/FunASR into dev_cmz2

Browse Source 
add
This commit is contained in:
游雁 2023-03-31 15:06:38 +08:00
commit d194b4bcd4
1 changed files with 2 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
funasr/export/models/encoder/sanm_encoder.py
View File

@ -164,6 +164,7 @@ class SANMVadEncoder(nn.Module):
speech = speech * self._output_size ** 0.5
mask = self.make_pad_mask(speech_lengths)
mask = self.prepare_mask(mask, sub_masks)
vad_mask = self.prepare_mask(mask, vad_mask)
if self.embed is None:
xs_pad = speech
else:
@ -175,7 +176,7 @@ class SANMVadEncoder(nn.Module):
# encoder_outs = self.model.encoders(xs_pad, mask)
for layer_idx, encoder_layer in enumerate(self.model.encoders):
if layer_idx == len(self.model.encoders) - 1:
mask = (mask[0], vad_mask)
mask = vad_mask
encoder_outs = encoder_layer(xs_pad, mask)
xs_pad, masks = encoder_outs[0], encoder_outs[1]