Skip to content

Fix VAE spatial sharding dynamic calculation bug in Wan pipeline. #407

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
ninatu wants to merge 1 commit into
base: main
Choose a base branch
from
+11 −3
Open

Fix VAE spatial sharding dynamic calculation bug in Wan pipeline. #407

Show file tree
Hide file tree
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
10 changes: 9 additions & 1 deletion src/maxdiffusion/pipelines/wan/wan_pipeline.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -676,7 +676,15 @@ def _create_common_components(cls, config, vae_only=False, i2v=False):
dp_size = mesh.shape.get("data", 1)
if dp_size == -1 or dp_size == 0:
dp_size = 1
vae_spatial = (2 * total_devices) // dp_size
non_dp_size = total_devices // dp_size
# VAE activations are huge; we want to maximize VAE spatial sharding (vae_spatial).
# Ideally, we want 2x the Transformer's model parallel size (2 * non_dp_size).
# However, vae_spatial must divide total_devices, which requires dp_size to be even.
# If dp_size is odd (e.g., single-device runs), we fallback to 1x (non_dp_size).
if dp_size % 2 == 0:
vae_spatial = 2 * non_dp_size
else:
vae_spatial = non_dp_size

assert (
total_devices % vae_spatial == 0
Expand Down
4 changes: 2 additions & 2 deletions src/maxdiffusion/pyconfig.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -281,8 +281,8 @@ def user_init(raw_keys):
raw_keys["global_batch_size_to_train_on"],
) = _HyperParameters.calculate_global_batch_sizes(raw_keys["per_device_batch_size"])

if raw_keys.get("vae_spatial", -1) == -1:
raw_keys["vae_spatial"] = 1
if "vae_spatial" not in raw_keys:
raw_keys["vae_spatial"] = -1


def get_num_slices(raw_keys):
Expand Down
Loading