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

Cast to >=float32 tensor when passing scalar to self.log #19046

Merged
merged 7 commits into from
Nov 24, 2023
Merged

Cast to >=float32 tensor when passing scalar to self.log #19046

Changes from 5 commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
2951ba2
fix: __to_tensor_high_precision
MF-FOOM Nov 10, 2023
0277e7c
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Nov 21, 2023
1c24b86
fix: self._get_default_high_precision_dtype()
MF-FOOM Nov 21, 2023
b9894c9
fix: make __get_default_high_precision_dtype a staticmethod
MF-FOOM Nov 21, 2023
7957008
refactor: __to_tensor_high_precision -> __to_tensor
MF-FOOM Nov 22, 2023
2d19b05
refactor: import _get_default_dtype from logger connector
MF-FOOM Nov 23, 2023
3ee07db
chlog
awaelchli Nov 24, 2023
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
12 changes: 11 additions & 1 deletion src/lightning/pytorch/core/module.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -621,8 +621,18 @@ def __check_not_nested(value: dict, name: str) -> None:
def __check_allowed(v: Any, name: str, value: Any) -> None:
raise ValueError(f"`self.log({name}, {value})` was called, but `{type(v).__name__}` values cannot be logged")

@staticmethod
MF-FOOM marked this conversation as resolved.
Show resolved Hide resolved
def __get_default_high_precision_dtype() -> torch.dtype:
"""The default dtype for new tensors, but no lower than float32."""
dtype = torch.get_default_dtype()
return dtype if dtype in (torch.float32, torch.float64) else torch.float32

def __to_tensor(self, value: Union[Tensor, numbers.Number], name: str) -> Tensor:
value = value.clone().detach() if isinstance(value, Tensor) else torch.tensor(value, device=self.device)
value = (
value.clone().detach()
if isinstance(value, Tensor)
else torch.tensor(value, device=self.device, dtype=self.__get_default_high_precision_dtype())
)
if not torch.numel(value) == 1:
raise ValueError(
f"`self.log({name}, {value})` was called, but the tensor must have a single element."
Expand Down
Loading