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

Jump to bottom

Add scaling option to loftq #2073

Closed
wants to merge 1 commit into from
Closed

Add scaling option to loftq #2073

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
1 change: 1 addition & 0 deletions src/peft/tuners/lora/layer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -194,6 +194,7 @@ def loftq_init(self, adapter_name):
"num_bits": self.kwargs.get("loftq_bits", 4),
"reduced_rank": self.r[adapter_name],
"num_iter": self.kwargs.get("loftq_iter", 1),
"scaling": self.scaling[adapter_name]
}

qweight, lora_A, lora_B = loftq_init(weight, **kwargs)
Expand Down
4 changes: 2 additions & 2 deletions src/peft/utils/loftq_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -187,7 +187,7 @@ def _low_rank_decomposition(weight, reduced_rank=32):


@torch.no_grad()
def loftq_init(weight: Union[torch.Tensor, torch.nn.Parameter], num_bits: int, reduced_rank: int, num_iter=1):
def loftq_init(weight: Union[torch.Tensor, torch.nn.Parameter], num_bits: int, reduced_rank: int, num_iter=1, scaling=1):
if is_bnb_available():
import bitsandbytes as bnb
else:
Expand Down Expand Up @@ -233,7 +233,7 @@ def loftq_init(weight: Union[torch.Tensor, torch.nn.Parameter], num_bits: int, r
L, R, reduced_rank = output["L"], output["R"], output["reduced_rank"]
res = weight - torch.mm(L, R)

lora_A, lora_B = R, L
lora_A, lora_B = R, L/scaling

return dequantized_weight.to(device=device, dtype=dtype), lora_A, lora_B

Expand Down