Multiply min_abs values in line of encoder residuals by 4.

This commit is contained in:
Daniel Povey 2022-12-29 12:49:04 +08:00
parent 71d7843654
commit 03e1f7dc01

View File

@ -461,7 +461,7 @@ class ZipformerEncoderLayer(nn.Module):
self.balancer = ActivationBalancer( self.balancer = ActivationBalancer(
embed_dim, channel_dim=-1, embed_dim, channel_dim=-1,
min_positive=0.45, max_positive=0.55, min_positive=0.45, max_positive=0.55,
min_abs=0.25, max_abs=6.0, min_abs=1.0, max_abs=6.0,
) )
self.whiten = Whiten(num_groups=1, self.whiten = Whiten(num_groups=1,
whitening_limit=_whitening_schedule(4.0, ratio=3.0), whitening_limit=_whitening_schedule(4.0, ratio=3.0),
@ -1478,7 +1478,7 @@ class NonlinAttentionModule(nn.Module):
self.balancer2 = ActivationBalancer( self.balancer2 = ActivationBalancer(
channels, channel_dim=-1, channels, channel_dim=-1,
min_positive=0.3, max_positive=0.7, min_positive=0.3, max_positive=0.7,
min_abs=ScheduledFloat((0.0, 0.001), (4000.0, 0.005)), min_abs=ScheduledFloat((0.0, 0.004), (4000.0, 0.02)),
prob=0.05, # out of concern for memory usage prob=0.05, # out of concern for memory usage
) )
@ -1714,7 +1714,7 @@ class ConvNeXt(nn.Module):
self.out_balancer = ActivationBalancer( self.out_balancer = ActivationBalancer(
channels, channel_dim=1, channels, channel_dim=1,
min_positive=0.4, max_positive=0.6, min_positive=0.4, max_positive=0.6,
min_abs=0.25, max_abs=6.0, min_abs=1.0, max_abs=6.0,
) )
self.out_whiten = Whiten(num_groups=1, self.out_whiten = Whiten(num_groups=1,
whitening_limit=5.0, whitening_limit=5.0,