Skip to content

Special handling of Torch DDP in callback #21081

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
merged 4 commits into from
Mar 22, 2025
Merged

Special handling of Torch DDP in callback #21081

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
8d34f6d
Special handling of Torch DDP in callback
timsweeneyfanelli Mar 21, 2025
ca30b2f
Use inheritance tree for DDP check
timsweeneyfanelli Mar 21, 2025
045cebd
Fixing comment.
timsweeneyfanelli Mar 21, 2025
cb50185
inlining DDP import where its needed.
timsweeneyfanelli Mar 21, 2025
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
13 changes: 13 additions & 0 deletions keras/src/callbacks/callback.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -76,6 +76,19 @@ def set_model(self, model):

@property
def model(self):
if backend.backend() == "torch":
from torch.nn.parallel import DistributedDataParallel

if isinstance(self._model, DistributedDataParallel):
# Keras Callbacks expect to work with Keras models. e.g
# ModelCheckpoint and EarlyStopping both attempt to call
# keras-specific APIs on the value returned from this
# property. If this callback was created against a DDP
# wrapper instead of the underlying keras.Model, it is
# likely to fail. Return self._model.module for DDP
# instances instead.
return self._model.module

if backend.backend() == "jax" and hasattr(
self._model, "jax_state_sync"
):
Expand Down