Skip to content

Fix in-place input normalization when using sentence/speaker norm #2504

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Adel-Moumen merged 1 commit into from
Apr 24, 2024
Merged

Fix in-place input normalization when using sentence/speaker norm #2504

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 12 additions & 5 deletions speechbrain/processing/features.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1082,6 +1082,11 @@ def forward(self, x, lengths, spk_ids=torch.tensor([]), epoch=0):
current_means = []
current_stds = []

if self.norm_type == "sentence" or self.norm_type == "speaker":
# we will do in-place slice assignments over `out`
out = torch.empty_like(x)
# otherwise don't assign it yet

for snt_id in range(N_batches):
# Avoiding padded time steps
actual_size = torch.round(lengths[snt_id] * x.shape[1]).int()
Expand All @@ -1095,7 +1100,7 @@ def forward(self, x, lengths, spk_ids=torch.tensor([]), epoch=0):
current_stds.append(current_std)

if self.norm_type == "sentence":
x[snt_id] = (x[snt_id] - current_mean.data) / current_std.data
out[snt_id] = (x[snt_id] - current_mean.data) / current_std.data

if self.norm_type == "speaker":
spk_id = int(spk_ids[snt_id][0])
Expand Down Expand Up @@ -1141,14 +1146,14 @@ def forward(self, x, lengths, spk_ids=torch.tensor([]), epoch=0):
speaker_mean = current_mean.data
speaker_std = current_std.data

x[snt_id] = (x[snt_id] - speaker_mean) / speaker_std
out[snt_id] = (x[snt_id] - speaker_mean) / speaker_std

if self.norm_type == "batch" or self.norm_type == "global":
current_mean = torch.mean(torch.stack(current_means), dim=0)
current_std = torch.mean(torch.stack(current_stds), dim=0)

if self.norm_type == "batch":
x = (x - current_mean.data) / (current_std.data)
out = (x - current_mean.data) / (current_std.data)

if self.norm_type == "global":
if self.training:
Expand All @@ -1175,9 +1180,11 @@ def forward(self, x, lengths, spk_ids=torch.tensor([]), epoch=0):

self.count = self.count + 1

x = (x - self.glob_mean.data.to(x)) / (self.glob_std.data.to(x))
out = (x - self.glob_mean.data.to(x)) / (
self.glob_std.data.to(x)
)

return x
return out

def _compute_current_stats(self, x):
"""Computes mean and std
Expand Down