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

Increase min_abs of balancer in NonlinAttentionModule from 1.5 to 2.0.

Browse Source
This commit is contained in:
Daniel Povey 2022-11-28 11:35:00 +08:00
parent 39ce60bb7c
commit c6d859dd05
1 changed files with 5 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
egs/librispeech/ASR/pruned_transducer_stateless7/zipformer.py
View File

@ -1433,12 +1433,15 @@ class NonlinAttentionModule(nn.Module):
self.in_proj = nn.Linear(channels, channels, bias=True) self.in_proj = nn.Linear(channels, channels, bias=True)
# balancer that goes before the sigmoid. # balancer that goes before the sigmoid. Have quite a large min_abs value, at 2.0,
# because we noticed that well-trained instances of this module have abs-value before the sigmoid
# starting from about 3, and poorly-trained instances of the module have smaller abs values
# before the sigmoid.
self.balancer = ActivationBalancer( self.balancer = ActivationBalancer(
channels // 2, channel_dim=-1, channels // 2, channel_dim=-1,
min_positive=ScheduledFloat((0.0, 0.1), (8000.0, 0.05)), min_positive=ScheduledFloat((0.0, 0.1), (8000.0, 0.05)),
max_positive=1.0, max_positive=1.0,
min_abs=1.5, min_abs=2.0,
max_abs=ScheduledFloat((0.0, 5.0), (8000.0, 10.0), default=1.0), max_abs=ScheduledFloat((0.0, 5.0), (8000.0, 10.0), default=1.0),
) )
self.sigmoid = nn.Sigmoid() self.sigmoid = nn.Sigmoid()