mrq/DL-Art-School
1
3
Fork 5
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Make feature encoder a bit more descriptive

Browse Source
This commit is contained in:
James Betker 2022-05-17 18:14:52 -06:00
parent ee364f4eeb
commit 9a9c3cafba
1 changed files with 3 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
codes/models/audio/mel2vec.py
View File

@ -366,8 +366,9 @@ class Mel2Vec(nn.Module):
self.input_blocks = nn.Sequential(nn.Conv1d(mel_input_channels, inner_dim//2, kernel_size=5, padding=2, stride=2),
nn.GroupNorm(num_groups=8, num_channels=inner_dim//2, affine=True),
nn.SiLU(),
nn.Conv1d(inner_dim//2, inner_dim, kernel_size=3, padding=1, stride=2),
nn.GroupNorm(num_groups=8, num_channels=inner_dim, affine=True),
nn.Conv1d(inner_dim//2, inner_dim, kernel_size=3, padding=1, stride=2),
nn.SiLU(),
nn.Conv1d(inner_dim, inner_dim, kernel_size=3, padding=1),
nn.SiLU(),
)
self.projector = Mel2Vec2FeatureProjection(inner_dim, dropout)