mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-09-08 16:44:20 +00:00
Code Issues Packages Projects Releases Wiki Activity

Change dropout_rate from 0.2 to 0.1; fix logging statement; fix assignment to rand_scales, nonrand_scales to use [:]

Browse Source
This commit is contained in:
Daniel Povey 2022-06-08 00:42:04 +08:00
parent a6050cb2de
commit 135be1e19c
2 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
egs/librispeech/ASR/pruned_transducer_stateless2/scaling.py
View File

@ -786,7 +786,7 @@ class Decorrelate(torch.nn.Module):
U, S, _ = norm_cov.svd()
if random.random() < 0.1:
print("Decorrelate: max,min eig of normalized cov is: {S.max().item():.2e},{S.min().item():.2e}")
logging.info(f"Decorrelate: max,min eig of normalized cov is: {S.max().item():.2e},{S.min().item():.2e}")
# row indexes of U correspond to channels, column indexes correspond to
# singular values: cov = U * diag(S) * U.t() where * is matmul.
@ -817,8 +817,8 @@ class Decorrelate(torch.nn.Module):
# rand_proportion is viewed as representing a proportion of the covariance, since
# the random and nonrandom components will not be correlated.
self.rand_scales = rand_proportion.sqrt()
self.nonrand_scales = (1.0 - rand_proportion).sqrt()
self.rand_scales[:] = rand_proportion.sqrt()
self.nonrand_scales[:] = (1.0 - rand_proportion).sqrt()
if True:

2
egs/librispeech/ASR/pruned_transducer_stateless5/conformer.py
View File

@ -199,7 +199,7 @@ class ConformerEncoderLayer(nn.Module):
)
self.dropout = torch.nn.Dropout(dropout)
self.decorrelate = Decorrelate(d_model, apply_prob=0.25, dropout_rate=0.2)
self.decorrelate = Decorrelate(d_model, apply_prob=0.25, dropout_rate=0.1)
def forward(