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

Fix dtype casting on CPU #339

Merged
merged 2 commits into from
Oct 26, 2023
Merged

Fix dtype casting on CPU #339

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6773d72
Fix dtype casting on CPU
epwalsh Oct 25, 2023
ce38e42
Merge branch 'main' into epwalsh/cpu-autocast
dirkgr Oct 26, 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
33 changes: 16 additions & 17 deletions olmo/model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -127,16 +127,15 @@ def build(cls, config: ModelConfig, size: Optional[int] = None, **kwargs) -> Lay
raise NotImplementedError(f"Not sure how to handle '{config.layer_norm_type}' LayerNorm type")

def _cast_if_autocast_enabled(self, tensor: torch.Tensor, dtype: Optional[torch.dtype] = None) -> torch.Tensor:
if torch.is_autocast_enabled():
if dtype is None:
if tensor.device.type == "cuda":
dtype = torch.get_autocast_gpu_dtype()
elif tensor.device.type == "cpu":
dtype = torch.get_autocast_cpu_dtype()
else:
raise NotImplementedError()
return tensor.to(dtype=dtype)
return tensor
# NOTE: `is_autocast_enabled()` only checks for CUDA autocast, so we use the separate function
# `is_autocast_cpu_enabled()` for CPU autocast.
# See https://github.com/pytorch/pytorch/issues/110966.
if tensor.device.type == "cuda" and torch.is_autocast_enabled():
return tensor.to(dtype=dtype if dtype is not None else torch.get_autocast_gpu_dtype())
elif tensor.device.type == "cpu" and torch.is_autocast_cpu_enabled():
return tensor.to(dtype=dtype if dtype is not None else torch.get_autocast_cpu_dtype())
else:
return tensor

def reset_parameters(self):
if self.weight is not None:
Expand Down Expand Up @@ -442,13 +441,13 @@ def reset_parameters(self):
@classmethod
def _cast_attn_bias(cls, bias: torch.Tensor, input_dtype: torch.dtype) -> torch.Tensor:
target_dtype = input_dtype
if torch.is_autocast_enabled():
if bias.device.type == "cuda":
target_dtype = torch.get_autocast_gpu_dtype()
elif bias.device.type == "cpu":
target_dtype = torch.get_autocast_cpu_dtype()
else:
raise NotImplementedError()
# NOTE: `is_autocast_enabled()` only checks for CUDA autocast, so we use the separate function
# `is_autocast_cpu_enabled()` for CPU autocast.
# See https://github.com/pytorch/pytorch/issues/110966.
if bias.device.type == "cuda" and torch.is_autocast_enabled():
target_dtype = torch.get_autocast_gpu_dtype()
elif bias.device.type == "cpu" and torch.is_autocast_cpu_enabled():
target_dtype = torch.get_autocast_cpu_dtype()
if bias.dtype != target_dtype:
bias = bias.to(target_dtype)
ensure_finite_(bias, check_neg_inf=True, check_pos_inf=False)
Expand Down
Loading