Skip to content

Refactor LoRAFeedForward.forward to use shared lora_call (#20831) #20831

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
meta-codesync merged 1 commit into from
Jul 10, 2026
+6 −13
Merged

Refactor LoRAFeedForward.forward to use shared lora_call (#20831) #20831

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
19 changes: 6 additions & 13 deletions examples/models/llama/feed_forward.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
import torch.nn.functional as F

from executorch.examples.models.llama.lora import LoRALinear
from executorch.examples.models.llama.lora import lora_call, LoRALinear
from executorch.examples.models.llama.model_args import ModelArgs
from torch import nn

Expand Down Expand Up @@ -65,15 +65,8 @@ def __init__(self, dim: int, hidden_dim: int, args: ModelArgs):
)

def forward(self, x, lora_blob=None):
# CoreML LoRA-as-IO Path-2: when `lora_blob` is provided, route per-
# projection slices to LoRALinear instances tagged with `_lora_key`.
# Default behavior (lora_blob=None) is unchanged.
def _call(linear, x_in):
if lora_blob is not None:
key = getattr(linear, "_lora_key", None)
if key is not None and key in lora_blob:
a, b = lora_blob[key]
return linear(x_in, a, b)
return linear(x_in)

return _call(self.w2, F.silu(_call(self.w1, x)) * _call(self.w3, x))
return lora_call(
self.w2,
F.silu(lora_call(self.w1, x, lora_blob)) * lora_call(self.w3, x, lora_blob),
lora_blob,
)
Loading