wassname/pytorch-ts
1
0
Fork 0
mirror of https://github.com/wassname/pytorch-ts.git synced 2026-08-04 13:13:57 +08:00
Code Issues Packages Projects Releases Wiki Activity

restore the original conv. Originally along target dim, but now time dim.

Browse Source
This commit is contained in:
wassname
2022-12-23 14:43:15 +08:00
parent 3fb53f620e
commit c8ab7d47df
1 changed files with 11 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files
pts/model/time_grad2/epsilon_theta.py
+11 -5
View File
@@ -44,7 +44,7 @@ class ResidualBlock(nn.Module):
)
self.diffusion_projection = nn.Linear(t_in, residual_channels)
self.conditioner_projection = nn.Conv1d(
cond_in, 2 * residual_channels, 1 #3, padding_mode="circular",
cond_in, 2 * residual_channels, 1, padding=2, padding_mode="circular",
)
self.output_projection = nn.Conv1d(residual_channels, 2 * residual_channels, 1)
@@ -94,7 +94,7 @@ class EpsilonTheta2(nn.Module):
):
super().__init__()
self.input_projection = nn.Conv1d(
target_dim, residual_channels, 1#, padding=2, padding_mode="circular"
target_dim, residual_channels, 1, padding=2, padding_mode="circular"
)
self.diffusion_embedding = DiffusionEmbedding(
time_emb_dim, proj_dim=residual_hidden
@@ -116,12 +116,18 @@ class EpsilonTheta2(nn.Module):
for i in range(residual_layers)
]
)
self.skip_projection = nn.Conv1d(residual_channels, residual_channels, 1)
self.output_projection = nn.Conv1d(residual_channels, target_dim, 1)
self.skip_projection = nn.Conv1d(residual_channels, residual_channels, 3)
# self.output_projection = nn.Conv1d(residual_channels, target_dim, 3)
nn.init.kaiming_normal_(self.input_projection.weight)
nn.init.kaiming_normal_(self.skip_projection.weight)
nn.init.zeros_(self.output_projection.weight)
# nn.init.zeros_(self.output_projection.weight)
self.output_projection = nn.Sequential(
nn.Conv1d(residual_channels, residual_channels, 3),
# nn.LeakyReLU(),
# nn.Conv1d(residual_channels, target_dim, 3, padding="same"),
)
def forward(self, inputs, time, cond):
x = self.input_projection(inputs)