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

previous_dtype is now inferred from F.linear's result output type. #1010

Merged
merged 16 commits into from
Feb 19, 2024
Merged

previous_dtype is now inferred from F.linear's result output type. #1010

Changes from 2 commits
Commits
Show all changes
16 commits
Select commit Hold shift + click to select a range
ec29aa8
previous_dtype is now inferred from F.linear's result output type.
Oct 10, 2023
552935a
make style formatting
Oct 11, 2023
2c1c8b0
merged with recent changes
Dec 6, 2023
f6ad119
added fix, also for Conv2D and Embedding layer
Dec 6, 2023
aac3781
added autocast compatibility test
Dec 6, 2023
0f0daae
run make style on test
Dec 6, 2023
96d9ec7
removed labels
Dec 15, 2023
3c7eb2e
added gpu requirement for this test
Dec 18, 2023
c09fbe7
review changes
Jan 24, 2024
0b4001c
merge with main
Jan 24, 2024
1748f21
added skip test
Jan 24, 2024
57f6d8d
reverted comment position
Jan 24, 2024
6d1df05
added comment for lora dtype consistency test
Feb 14, 2024
6b38805
Merge remote-tracking branch 'peft/main'
Feb 19, 2024
c697dbe
use regular assert to comply with make style
Feb 19, 2024
751c5ce
make style
Feb 19, 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
4 changes: 3 additions & 1 deletion src/peft/tuners/lora/layer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -240,16 +240,18 @@ def _linear(self, input: torch.Tensor) -> torch.Tensor:
return F.linear(input, transpose(self.weight, self.fan_in_fan_out), bias=self.bias)

def forward(self, x: torch.Tensor) -> torch.Tensor:
previous_dtype = x.dtype

if self.disable_adapters:
if self.merged:
self.unmerge()
result = self._linear(x)
previous_dtype = result.dtype
elif self.merged:
result = self._linear(x)
previous_dtype = result.dtype
else:
result = self._linear(x)
previous_dtype = result.dtype
for active_adapter in self.active_adapters:
if active_adapter not in self.lora_A.keys():
continue
Expand Down