Skip to content

Fast ddp in prepare_model #2275

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
regisss merged 1 commit into from
Sep 22, 2025
+1 −2
Merged

Fast ddp in prepare_model #2275

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
3 changes: 1 addition & 2 deletions optimum/habana/accelerate/accelerator.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -116,7 +116,7 @@ def prepare_model(

if device_placement and not self.verify_device_map(model):
model = model.to(self.device)
if not evaluation_mode:
if not evaluation_mode and self.distribution_strategy != "fast_ddp":
if self.multi_device and not (self.parallelism_config and self.parallelism_config.tp_enabled):
if model_has_dtensor(model):
raise ValueError(
Expand Down Expand Up @@ -319,7 +319,6 @@ def prepare_model(
# model = super().prepare_model(model, device_placement=device_placement, evaluation_mode=True)
# else:
# model = super().prepare_model(model, device_placement=device_placement, evaluation_mode=evaluation_mode)

# return model

# INFO: this adds support for autograd compilation to the deepspeed engine
Expand Down
Loading