Skip to content

fix(wan): apply patchify flatten in forward/consumers (fixes #1063) #1435

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
genno-whittlery wants to merge 2 commits into
base: main
Choose a base branch
from
Open

fix(wan): apply patchify flatten in forward/consumers (fixes #1063) #1435

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
5d8a8bf
fix(wan): WanModel.patchify flatten + return grid size tuple
genno-whittlery May 11, 2026
ecb6f53
fix(wan): apply patchify fix in consumers, keep signature
genno-whittlery May 11, 2026
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 diffsynth/models/wan_video_dit.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -527,8 +527,10 @@ def forward(self,
clip_embdding = self.img_emb(clip_feature)
context = torch.cat([clip_embdding, context], dim=1)

x, (f, h, w) = self.patchify(x)

x = self.patchify(x)
f, h, w = x.shape[-3], x.shape[-2], x.shape[-1]
x = x.flatten(2).transpose(1, 2)

freqs = torch.cat([
self.freqs[0][:f].view(f, 1, 1, -1).expand(f, h, w, -1),
self.freqs[1][:h].view(1, h, 1, -1).expand(f, h, w, -1),
Expand Down
8 changes: 6 additions & 2 deletions diffsynth/pipelines/wan_video.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1658,11 +1658,15 @@ def model_fn_wans2v(

# x and s2v_pose_latents
s2v_pose_latents = torch.zeros_like(x) if s2v_pose_latents is None else s2v_pose_latents
x, (f, h, w) = dit.patchify(dit.patch_embedding(x) + dit.cond_encoder(s2v_pose_latents))
x = dit.patchify(dit.patch_embedding(x) + dit.cond_encoder(s2v_pose_latents))
f, h, w = x.shape[-3], x.shape[-2], x.shape[-1]
x = x.flatten(2).transpose(1, 2)
seq_len_x = seq_len_x_global = x.shape[1] # global used for unified sequence parallel

# reference image
ref_latents, (rf, rh, rw) = dit.patchify(dit.patch_embedding(origin_ref_latents))
ref_latents = dit.patchify(dit.patch_embedding(origin_ref_latents))
rf, rh, rw = ref_latents.shape[-3], ref_latents.shape[-2], ref_latents.shape[-1]
ref_latents = ref_latents.flatten(2).transpose(1, 2)
grid_sizes = dit.get_grid_sizes((f, h, w), (rf, rh, rw))
x = torch.cat([x, ref_latents], dim=1)
# mask
Expand Down
6 changes: 4 additions & 2 deletions diffsynth/utils/xfuser/xdit_context_parallel.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -81,8 +81,10 @@ def usp_dit_forward(self,
clip_embdding = self.img_emb(clip_feature)
context = torch.cat([clip_embdding, context], dim=1)

x, (f, h, w) = self.patchify(x)

x = self.patchify(x)
f, h, w = x.shape[-3], x.shape[-2], x.shape[-1]
x = x.flatten(2).transpose(1, 2)

freqs = torch.cat([
self.freqs[0][:f].view(f, 1, 1, -1).expand(f, h, w, -1),
self.freqs[1][:h].view(1, h, 1, -1).expand(f, h, w, -1),
Expand Down