Skip to content

[chore] raise as early as possible in group offloading #11792

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 2 commits into from
Jun 24, 2025
Merged

[chore] raise as early as possible in group offloading #11792

Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
0d1cd52
raise as early as possible in group offloading
sayakpaul Jun 24, 2025
c86e4b5
remove check from ModuleGroup
sayakpaul Jun 24, 2025
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
9 changes: 6 additions & 3 deletions src/diffusers/hooks/group_offloading.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,6 +63,9 @@ def __init__(
onload_self: bool = True,
offload_to_disk_path: Optional[str] = None,
) -> None:
if stream is None and record_stream:
Copy link
Contributor

Choose a reason for hiding this comment

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

We can remove this completely, no? It's not supposed to be user facing and now apply_group_offloading handles the error case too

Copy link
Member Author

Choose a reason for hiding this comment

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

We can but since ModuleGroup is a public class I thought of keeping it.

Copy link
Contributor

Choose a reason for hiding this comment

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

Not really, it's not a public class - no user is expected to import it or that we document it that way. We don't necessarily need to mark every internal data structure private to imply it shouldn't be used by users

Copy link
Member Author

Choose a reason for hiding this comment

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

Removed.

raise ValueError("`record_stream` cannot be True when `stream` is None.")

self.modules = modules
self.offload_device = offload_device
self.onload_device = onload_device
Expand Down Expand Up @@ -96,9 +99,6 @@ def __init__(
else:
self.cpu_param_dict = self._init_cpu_param_dict()

if self.stream is None and self.record_stream:
raise ValueError("`record_stream` cannot be True when `stream` is None.")

def _init_cpu_param_dict(self):
cpu_param_dict = {}
if self.stream is None:
Expand Down Expand Up @@ -513,6 +513,9 @@ def apply_group_offloading(
else:
raise ValueError("Using streams for data transfer requires a CUDA device, or an Intel XPU device.")

if not use_stream and record_stream:
raise ValueError("`record_stream` cannot be True when `use_stream=False`.")

_raise_error_if_accelerate_model_or_sequential_hook_present(module)

if offload_type == "block_level":
Expand Down