mrq/DL-Art-School
1
3
Fork 5
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Don't step if NaN losses are encountered.

Browse Source
This commit is contained in:
James Betker 2022-02-24 17:45:08 -07:00
parent ea500ad42a
commit 18dc62453f
1 changed files with 3 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
codes/trainer/steps.py
View File

@ -263,7 +263,7 @@ class ConfigurableStep(Module):
loss.clear_metrics()
# In some cases, the loss could not be set (e.g. all losses have 'after')
if train and isinstance(total_loss, torch.Tensor):
if train and isinstance(total_loss, torch.Tensor) and total_loss.isfinite():
loss_accumulator.add_loss("%s_total" % (self.get_training_network_name(),), total_loss)
reset_required = total_loss < self.min_total_loss
@ -282,6 +282,8 @@ class ConfigurableStep(Module):
loss_accumulator.increment_metric("%s_skipped_steps" % (self.get_training_network_name(),))
self.grads_generated = True
elif not total_loss.isfinite():
print("Non-finite loss encountered. Skipping backwards step.")
# Detach all state variables. Within the step, gradients can flow. Once these variables leave the step
# we must release the gradients.