Skip to content

[Bugfix][NPU][XPU] Use platform-aware profiler activities for trace generation #1542

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

Closed
lishunyang12 wants to merge 3 commits into from
+27 −1
Closed

[Bugfix][NPU][XPU] Use platform-aware profiler activities for trace generation #1542

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
f433b1a
[Bugfix][NPU][XPU] Use platform-aware profiler activities for trace g…
lishunyang12 Mar 4, 2026
754fd17
Merge branch 'main' into fix/npu-profiler-activity
lishunyang12 Mar 12, 2026
7390d6c
Merge branch 'main' into fix/npu-profiler-activity
lishunyang12 Mar 18, 2026
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
28 changes: 27 additions & 1 deletion vllm_omni/diffusion/profiler/torch_profiler.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,32 @@
logger = init_logger(__name__)


def _get_profiler_activities() -> list[ProfilerActivity]:
"""Return the appropriate profiler activities for the current platform."""
from vllm_omni.platforms import current_omni_platform

activities = [ProfilerActivity.CPU]
device_type = current_omni_platform.device_type
if device_type == "npu":

@david6666666 david6666666 Feb 28, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Does it support other platforms(rocm, xpu)? Does it require adaptation?

@lishunyang12 lishunyang12 Feb 28, 2026

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ROCm uses \ in PyTorch's profiler API, so the else branch already covers it. XPU support can be added when needed — this PR is scoped to fix #1484.

@lishunyang12 lishunyang12 Feb 28, 2026 •
edited
Loading

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Good point — I'll add XPU support too. ROCm already works with the CUDA fallback since PyTorch maps it to ProfilerActivity.CUDA.

# torch_npu registers ProfilerActivity.NPU on import
try:
import torch_npu # noqa: F401
except ImportError:
pass
if hasattr(ProfilerActivity, "NPU"):
activities.append(ProfilerActivity.NPU)
else:
logger.warning("ProfilerActivity.NPU not available. Make sure torch_npu is installed.")
elif device_type == "xpu":
if hasattr(ProfilerActivity, "XPU"):
activities.append(ProfilerActivity.XPU)
else:
logger.warning("ProfilerActivity.XPU not available. Make sure Intel XPU support is installed.")
else:
activities.append(ProfilerActivity.CUDA)
return activities


class TorchProfiler(ProfilerBase):
"""
Torch-based profiler configured for End-to-End continuous recording.
Expand Down Expand Up @@ -70,7 +96,7 @@ def trace_handler(p):

# 4. Initialize profiler with long active period
cls._profiler = profile(
activities=[ProfilerActivity.CPU, ProfilerActivity.CUDA],
activities=_get_profiler_activities(),
schedule=torch.profiler.schedule(
wait=0,
warmup=0,
Expand Down
Loading