Skip to content

Move buffers to device #10523

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 4 commits into from
Jan 16, 2025
Merged

Move buffers to device #10523

Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
570d811
Move buffers to device
hlky Jan 10, 2025
3d56e94
add test
hlky Jan 10, 2025
bb2e228
named_buffers
hlky Jan 14, 2025
f20d888
Merge branch 'main' into model-loading-buffers-device
hlky Jan 16, 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
11 changes: 11 additions & 0 deletions src/diffusers/models/model_loading_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -246,6 +246,17 @@ def load_model_dict_into_meta(
else:
set_module_tensor_to_device(model, param_name, device, value=param)

for param_name, param in model.named_buffers():
Copy link
Collaborator

Choose a reason for hiding this comment

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

can we move the changes outside of the function? it takes a state_dict as input so it is strange that it also do something with model.named_buffer()

would it work if we run load_model_dict_into_meta on the buffers in a seperate call? if not, we could add an additional argument

if is_quantized and (
hf_quantizer.check_if_quantized_param(model, param, param_name, state_dict, param_device=device)
):
hf_quantizer.create_quantized_param(model, param, param_name, device, state_dict, unexpected_keys)
else:
if accepts_dtype:
set_module_tensor_to_device(model, param_name, device, value=param, **set_module_kwargs)
else:
set_module_tensor_to_device(model, param_name, device, value=param)

Copy link
Member

Choose a reason for hiding this comment

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

Interesting. Let's also have a test for this?

Cc: @SunMarc too.

Copy link
Member

Choose a reason for hiding this comment

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

persistant buffer should be in the state dict no ? It is for non persistant buffers ?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Yes, for non-persistent buffers.

return unexpected_keys


Expand Down
Loading