Skip to content

plaintext Enhance dtype consistency checks in LoRALoaderMixin #11

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

Open
mandlinsarah wants to merge 1 commit into
base: main
Choose a base branch
from
Open

plaintext Enhance dtype consistency checks in LoRALoaderMixin #11

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
17 changes: 10 additions & 7 deletions src/mistral_inference/lora.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -102,12 +102,15 @@ def load_lora(self, lora_path: Union[Path, str], scaling: float = 2.0) -> None:

def _load_lora_state_dict(self, lora_state_dict: Dict[str, torch.Tensor], scaling: float = 2.0) -> None:
"""Loads LoRA state_dict"""
lora_dtypes = set([p.dtype for p in lora_state_dict.values()])
assert (
len(lora_dtypes) == 1
), f"LoRA weights have multiple different dtypes {lora_dtypes}. All weights need to have the same dtype"
lora_dtype = lora_dtypes.pop()
assert lora_dtype == self.dtype, f"LoRA weights dtype differs from model's dtype {lora_dtype} != {self.dtype}" # type: ignore[attr-defined]
lora_dtypes = set(p.dtype for p in lora_state_dict.values())
assert len(lora_dtypes) == 1, (
f"LoRA weights have multiple different dtypes {lora_dtypes}. "
"All weights need to have the same dtype"
)
lora_dtype = next(iter(lora_dtypes))
assert lora_dtype == self.dtype, (
f"LoRA weights dtype ({lora_dtype}) differs from model's dtype ({self.dtype})"
) # type: ignore[attr-defined]
assert all("lora" in key for key in lora_state_dict.keys())

# move tensors to device
Expand Down Expand Up @@ -152,4 +155,4 @@ def _load_lora_state_dict(self, lora_state_dict: Dict[str, torch.Tensor], scalin
self.pipeline_rank, # type: ignore[attr-defined]
)

self.load_state_dict(state_dict, strict=True) # type: ignore[attr-defined]
self.load_state_dict(state_dict, strict=True) # type: ignore[attr-defined]