Skip to content

[NaN check] Add NaN check to support bfloat16. #5879

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

Closed
wants to merge 4 commits into from
Closed

[NaN check] Add NaN check to support bfloat16. #5879

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
5acbfa5
add nan check to support bfloat16
ys950902 Aug 8, 2024
125e193
Merge branch 'master' into nan_check
tjruwase Aug 9, 2024
503e0b9
Merge branch 'master' into nan_check
tjruwase Sep 3, 2024
7f4d8c5
Merge branch 'master' into nan_check
tohtana Sep 27, 2024
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
2 changes: 1 addition & 1 deletion deepspeed/runtime/zero/stage3.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1997,7 +1997,7 @@ def _overflow_clean_up(self, prev_scale):
def _overflow_check_and_loss_scale_update(self):

# First compute norm for all group so we know if there is overflow
if self.dtype == torch.float16:
if self.dtype in [torch.float16, torch.bfloat16]:
self.check_overflow()

#loss scaling related computation
Expand Down
2 changes: 1 addition & 1 deletion deepspeed/runtime/zero/stage_1_and_2.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1828,7 +1828,7 @@ def step(self, closure=None):
see_memory_usage(f"In step before checking overflow")

# First compute norm for all group so we know if there is overflow
if self.dtype == torch.float16:
if self.dtype in [torch.float16, torch.bfloat16]:
self.check_overflow()

prev_scale = self.loss_scale
Expand Down
Loading