mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-12-11 06:55:27 +00:00
Code Issues Packages Projects Releases Wiki Activity

Set lr_factor on to_scores, max_abs=4.0 on balancer

Browse Source
This commit is contained in:
Daniel Povey 2023-05-23 10:55:33 +08:00
parent 45043e2e21
commit 3a71a53d8d
1 changed files with 3 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
egs/libriheavy/LM/zipformer1/subformer.py
View File

@ -813,6 +813,7 @@ class LearnedDownsamplingModule(nn.Module):
super().__init__()
self.to_scores = nn.Linear(embed_dim, 1, bias=False)
self.to_scores.lr_factor = 0.5
# score_balancer is just to keep the magnitudes of the scores in
# a fixed range and keep them balanced around zero, to stop
# these drifting around.
@ -820,7 +821,8 @@ class LearnedDownsamplingModule(nn.Module):
self.score_balancer = Balancer(1, channel_dim=-1,
min_positive=1/(2*downsampling_factor),
max_positive=0.6,
min_abs=1.0)
min_abs=1.0,
max_abs=4.0)
# below are for diagnostics.
self.copy_weights1 = nn.Identity()