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

Fix clamping of bypass scale; remove a couple unused variables.

Browse Source
This commit is contained in:
Daniel Povey 2022-10-27 14:05:53 +08:00
parent bf37c7ca85
commit 938510ac9f
1 changed files with 7 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
egs/librispeech/ASR/pruned_transducer_stateless7/conformer.py
View File

@ -360,13 +360,13 @@ class ConformerEncoderLayer(nn.Module):
delta = src - src_orig
bypass_scale = self.bypass_scale
if self.training and random.random() < 0.25:
# with probability 0.25, in training mode, clamp bypass_scale to [
# 0.1, 1.0 ]; this will encourage it to learn parameters within this
# range by making parameters that are outside that range range
# noisy.
if torch.jit.is_scripting() or (not self.training) or random.random() > 0.1:
# with probability 0.9, in training mode, or always, in testing
# mode, clamp bypass_scale to [ 0.1, 1.0 ]; this will encourage it
# to learn parameters within this range by making parameters that
# are outside that range range noisy.
bypass_scale = bypass_scale.clamp(min=0.1, max=1.0)
src = src_orig + delta * self.bypass_scale
src = src_orig + delta * bypass_scale
return self.whiten(src)
@ -532,8 +532,6 @@ class ConformerEncoder(nn.Module):
pos_emb = self.encoder_pos(src)
output = src
outputs = []
rnd_seed = src.numel() + random.randint(0, 1000)
layers_to_drop = self.get_layers_to_drop(rnd_seed, self.get_warmup_count())
@ -650,7 +648,7 @@ class AttentionDownsample(torch.nn.Module):
(seq_len, batch_size, in_channels) = src.shape
ds = self.downsample
d_seq_len = (seq_len + ds - 1) // ds
src_orig = src
# Pad to an exact multiple of self.downsample
if seq_len != d_seq_len * ds:
# right-pad src, repeating the last element.