Skip to content

Use float32 RoPE freqs in Wan with MPS backends #11643

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
merged 1 commit into from
Jun 2, 2025
+4 −2
Merged

Use float32 RoPE freqs in Wan with MPS backends #11643

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
6 changes: 4 additions & 2 deletions src/diffusers/models/transformers/transformer_wan.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -72,7 +72,8 @@ def __call__(
if rotary_emb is not None:

def apply_rotary_emb(hidden_states: torch.Tensor, freqs: torch.Tensor):
x_rotated = torch.view_as_complex(hidden_states.to(torch.float64).unflatten(3, (-1, 2)))
dtype = torch.float32 if hidden_states.device.type == "mps" else torch.float64
x_rotated = torch.view_as_complex(hidden_states.to(dtype).unflatten(3, (-1, 2)))
x_out = torch.view_as_real(x_rotated * freqs).flatten(3, 4)
return x_out.type_as(hidden_states)

Expand Down Expand Up @@ -190,9 +191,10 @@ def __init__(
t_dim = attention_head_dim - h_dim - w_dim

freqs = []
freqs_dtype = torch.float32 if torch.backends.mps.is_available() else torch.float64
for dim in [t_dim, h_dim, w_dim]:
freq = get_1d_rotary_pos_embed(
dim, max_seq_len, theta, use_real=False, repeat_interleave_real=False, freqs_dtype=torch.float64
dim, max_seq_len, theta, use_real=False, repeat_interleave_real=False, freqs_dtype=freqs_dtype
)
freqs.append(freq)
self.freqs = torch.cat(freqs, dim=1)
Expand Down