Skip to content
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

Make LoRACompatibleConv padding_mode work. #6031

Merged
merged 9 commits into from
Feb 27, 2024
Merged

Make LoRACompatibleConv padding_mode work. #6031

Show file tree
Hide file tree
Changes from 7 commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
2994616
Make LoRACompatibleConv padding_mode work.
jinghuan-Chen Dec 3, 2023
d5ea454
Format code style.
jinghuan-Chen Dec 4, 2023
c2de7c7
add fast test
jinghuan-Chen Dec 5, 2023
234a405
Update src/diffusers/models/lora.py
jinghuan-Chen Jan 7, 2024
a441798
code refactor
jinghuan-Chen Jan 10, 2024
5dcc27d
apply patrickvonplaten suggestion to simplify the code.
jinghuan-Chen Feb 11, 2024
949b01e
rm test_lora_layers_old_backend.py and add test case in test_lora_lay…
jinghuan-Chen Feb 24, 2024
f7ef216
update test case.
jinghuan-Chen Feb 25, 2024
87ec26a
Merge branch 'main' into make-LoRACompatibleConv-paddingmode-work
yiyixuxu Feb 26, 2024
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
19 changes: 11 additions & 8 deletions src/diffusers/models/lora.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -352,16 +352,19 @@ def _unfuse_lora(self):
self.w_down = None

def forward(self, hidden_states: torch.Tensor, scale: float = 1.0) -> torch.Tensor:
if self.padding_mode != "zeros":
hidden_states = F.pad(hidden_states, self._reversed_padding_repeated_twice, mode=self.padding_mode)
padding = (0, 0)
else:
padding = self.padding

original_outputs = F.conv2d(
hidden_states, self.weight, self.bias, self.stride, padding, self.dilation, self.groups
)

if self.lora_layer is None:
# make sure to the functional Conv2D function as otherwise torch.compile's graph will break
# see: https://github.com/huggingface/diffusers/pull/4315
return F.conv2d(
hidden_states, self.weight, self.bias, self.stride, self.padding, self.dilation, self.groups
)
return original_outputs
else:
original_outputs = F.conv2d(
hidden_states, self.weight, self.bias, self.stride, self.padding, self.dilation, self.groups
)
return original_outputs + (scale * self.lora_layer(hidden_states))


Expand Down
Loading
Loading