Skip to content

Fix checker of nan/inf #9029

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
wawltor merged 2 commits into from
Aug 29, 2024
Merged

Fix checker of nan/inf #9029

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
276c477
fix checker position for nan/inf
ForFishes Aug 28, 2024
0a0ebcf
fix checker position for nan/inf
ForFishes Aug 28, 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
10 changes: 5 additions & 5 deletions paddlenlp/trainer/trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -992,10 +992,6 @@ def _inner_training_loop(
else:
tr_loss_step = self.training_step(model, inputs)

if not args.fp16:
if not paddle.isfinite(tr_loss_step).all().item():
raise ValueError(f"Loss contains inf or nan values at rank {paddle.distributed.get_rank()}")

tr_loss += tr_loss_step

def fused_allreduce_gradients_no_sync(paramlist, hcg):
Expand Down Expand Up @@ -1294,7 +1290,11 @@ def _print_timer(self):

def _get_item_from_loss(self, loss):
assert isinstance(loss, paddle.Tensor) and loss._is_initialized()
return loss.item()
loss_value = loss.item()
if not self.args.fp16:
if not np.isfinite(loss_value).all():
raise ValueError(f"Loss contains inf or nan values, its value is {loss_value}")
return loss_value

def _maybe_log_save_evaluate(self, tr_loss, model, epoch, ignore_keys_for_eval, **kwargs):
if self.control.should_log:
Expand Down