Skip to content

Fix timestep embedding dim in UNet1D #6692

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

Open
AndreasBergmeister wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Fix timestep embedding dim in UNet1D #6692

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
14 changes: 8 additions & 6 deletions src/diffusers/models/unets/unet_1d.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -99,7 +99,7 @@ def __init__(
# time
if time_embedding_type == "fourier":
self.time_proj = GaussianFourierProjection(
embedding_size=8, set_W_to_weight=False, log=False, flip_sin_to_cos=flip_sin_to_cos
embedding_size=block_out_channels[0], set_W_to_weight=False, log=False, flip_sin_to_cos=flip_sin_to_cos
Copy link
Copy Markdown
Collaborator

@yiyixuxu yiyixuxu Feb 23, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this is a breaking change

)
timestep_input_dim = 2 * block_out_channels[0]
elif time_embedding_type == "positional":
Expand All @@ -114,8 +114,10 @@ def __init__(
in_channels=timestep_input_dim,
time_embed_dim=time_embed_dim,
act_fn=act_fn,
out_dim=block_out_channels[0],
out_dim=time_embed_dim,
)
else:
time_embed_dim = timestep_input_dim

self.down_blocks = nn.ModuleList([])
self.mid_block = None
Expand All @@ -138,7 +140,7 @@ def __init__(
num_layers=layers_per_block,
in_channels=input_channel,
out_channels=output_channel,
temb_channels=block_out_channels[0],
temb_channels=time_embed_dim,
add_downsample=not is_final_block or downsample_each_block,
)
self.down_blocks.append(down_block)
Expand All @@ -149,7 +151,7 @@ def __init__(
in_channels=block_out_channels[-1],
mid_channels=block_out_channels[-1],
out_channels=block_out_channels[-1],
embed_dim=block_out_channels[0],
embed_dim=time_embed_dim,
num_layers=layers_per_block,
add_downsample=downsample_each_block,
)
Expand All @@ -175,7 +177,7 @@ def __init__(
num_layers=layers_per_block,
in_channels=prev_output_channel,
out_channels=output_channel,
temb_channels=block_out_channels[0],
temb_channels=time_embed_dim,
add_upsample=not is_final_block,
)
self.up_blocks.append(up_block)
Expand All @@ -186,7 +188,7 @@ def __init__(
self.out_block = get_out_block(
out_block_type=out_block_type,
num_groups_out=num_groups_out,
embed_dim=block_out_channels[0],
embed_dim=time_embed_dim,
out_channels=out_channels,
act_fn=act_fn,
fc_dim=block_out_channels[-1] // 4,
Expand Down