A bug in AdaLayerNorm

[akstt]

Describe the bug

diffusers/blob/main/src/diffusers/models/normalization.py The default value of dim parameter of chunk function is 0. The shape of scale and the shape of x doesn't match.

Reproduction

The shape of scale and the shape of x doesn't match.

Logs

No response

System Info

This is still in the main branch.

Who can help?

No response

[sayakpaul]

Reproducible code snippet please.

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.

[sayakpaul]

Closing because of inactivity.

[mohamedhassanmus]

Here is a reproducable code

import torch
from diffusers.models.normalization import AdaLayerNorm

bs = 32
dim = 100
num_steps = 1000
x = torch.randn(bs, dim)
t = torch.randint(0, num_steps, (bs,))
norm = AdaLayerNorm(dim, num_steps)
output = norm(x, t)

Log

  File "normalization.py", line 47, in forward
    x = self.norm(x) * (1 + scale) + shift
RuntimeError: The size of tensor a (100) must match the size of tensor b (200) at non-singleton dimension 1