wassname/TTS
1
0 Fork 0
mirror of https://github.com/wassname/TTS.git synced 2026-10-26 12:20:32 +08:00
Code Issues Packages Projects Releases Wiki Activity

larger attn loss

Browse Source
This commit is contained in:
Eren Golge committed 2018-04-24 13:21:10 -07:00
1 parent 85bbab74b3
commit 17ef3de1b9
1 file changed
+1 -1
Download Patch File Download Diff File Expand all files Collapse all files
train.py
+1 -1
View File
@@ -129,7 +129,7 @@ def train(model, criterion, data_loader, optimizer, epoch):
linear_spec_var[:, :, :n_priority_freq],
mel_lengths_var)
attention_loss = criterion(alignments, M, mel_lengths_var)
loss = mel_loss + linear_loss + 0.2 * attention_loss
loss = mel_loss + linear_loss + attention_loss
# backpass and check the grad norm
loss.backward()