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 pull request #796 from alibaba-damo-academy/dev_lhn

Browse Source 
fix uniasr finetune bug
This commit is contained in:
hnluo 2023-08-02 10:52:03 +08:00 committed by GitHub
commit 0b359919f1
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
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/modules/subsampling.py
View File

@ -358,7 +358,8 @@ class Conv1dSubsampling(torch.nn.Module):
"""
x = x.transpose(1, 2) # (b, d ,t)
x = self.pad_fn(x)
x = F.relu(self.conv(x))
#x = F.relu(self.conv(x))
x = F.leaky_relu(self.conv(x), negative_slope=0.)
x = x.transpose(1, 2) # (b, t ,d)
if x_len is None: