Skip to content

fix: Disable torch.compile for PI05 policy on MPS device #2723

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

Open
pranavsaroha wants to merge 2 commits into
base: main
Choose a base branch
from
Open

fix: Disable torch.compile for PI05 policy on MPS device #2723

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
74b389c
fix: Disable torch.compile for PI05 policy on MPS device
pranavsaroha Dec 26, 2025
e33f688
Merge branch 'main' into fix/pi05-mps-torch-compile-support
pkooij Dec 31, 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
15 changes: 11 additions & 4 deletions src/lerobot/policies/pi05/modeling_pi05.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -543,11 +543,18 @@ def __init__(self, config: PI05Config, rtc_processor: RTCProcessor | None = None
self.gradient_checkpointing_enabled = False

# Compile model if requested
# Note: torch.compile is not fully supported on MPS (Apple Silicon) yet
if config.compile_model:
torch.set_float32_matmul_precision("high")
self.sample_actions = torch.compile(self.sample_actions, mode=config.compile_mode)
# Also compile the main forward pass used during training
self.forward = torch.compile(self.forward, mode=config.compile_mode)
if config.device == "mps":
logging.warning(
"torch.compile is not fully supported on MPS device. "
"Skipping model compilation. Performance may be slower than on CUDA."
)
else:
torch.set_float32_matmul_precision("high")
self.sample_actions = torch.compile(self.sample_actions, mode=config.compile_mode)
# Also compile the main forward pass used during training
self.forward = torch.compile(self.forward, mode=config.compile_mode)

msg = """An incorrect transformer version is used, please create an issue on https://github.com/huggingface/lerobot/issues"""

Expand Down