[Bugfix][NPU][XPU] Use platform-aware profiler activities for trace generation#1542
Closed
lishunyang12 wants to merge 3 commits into
Closed
[Bugfix][NPU][XPU] Use platform-aware profiler activities for trace generation#1542lishunyang12 wants to merge 3 commits into
lishunyang12 wants to merge 3 commits into
Conversation
Collaborator
|
Thanks! I will test it on NPU. |
lishunyang12
force-pushed
the
fix/npu-profiler-activity
branch
from
352bb61 to
3b5b321
Compare
|
|
||
| activities = [ProfilerActivity.CPU] | ||
| device_type = current_omni_platform.device_type | ||
| if device_type == "npu": |
Collaborator
There was a problem hiding this comment.
Does it support other platforms(rocm, xpu)? Does it require adaptation?
Collaborator
Author
There was a problem hiding this comment.
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.
Collaborator
Author
There was a problem hiding this comment.
Good point — I'll add XPU support too. ROCm already works with the CUDA fallback since PyTorch maps it to ProfilerActivity.CUDA.
Collaborator
Author
|
@xuechendi PTAL |
lishunyang12
changed the title
xuechendi
reviewed
Feb 28, 2026
| activities.append(getattr(ProfilerActivity, "NPU")) | ||
| elif device_type == "xpu": | ||
| # Intel XPU support | ||
| activities.append(getattr(ProfilerActivity, "XPU")) |
Contributor
There was a problem hiding this comment.
This is how we did in vLLM main repo
TorchProfilerActivity = Literal["CPU", "CUDA", "XPU"]
TorchProfilerActivityMap = {
"CPU": torch.profiler.ProfilerActivity.CPU,
"CUDA": torch.profiler.ProfilerActivity.CUDA,
"XPU": torch.profiler.ProfilerActivity.XPU,
}
Current codes with getattr also works. Thanks for adding XPU
Collaborator
|
please fix DCO |
5 tasks
Contributor
|
Hi, I still encounter error |
lishunyang12
force-pushed
the
fix/npu-profiler-activity
branch
from
ac733e2 to
c9961c9
Compare
Collaborator
Author
|
@JustQJ Please tried it again. I made some changes according to your bug results. :) |
Contributor
Hi, In my test, my env |
ApsarasX
approved these changes
Mar 3, 2026
Collaborator
|
@lishunyang12 @JustQJ any progress？ |
…eneration Signed-off-by: lishunyang <lishunyang12@163.com>
lishunyang12
force-pushed
the
fix/npu-profiler-activity
branch
from
c9961c9 to
f433b1a
Compare
Collaborator
Author
|
@JustQJ The crash is fixed — the The reason @david6666666 DCO is fixed, squashed into one commit. Ready to go. |
Collaborator
Author
|
Should we accelerate this PR? @david6666666 @gcanlin |
Collaborator
|
@lishunyang12 Thanks for your contribution. #1261 is integrating NPU activities. NPU needs to use torch_npu api to get the activities after my check. Please take a look. I will add co-author in #1261 for you :) |
Collaborator
Author
Thanks @gcanlin |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
7 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
TorchProfilerhardcodesProfilerActivity.CUDA, which fails on NPU (Ascend) devices sinceCUDAactivity is not available there.current_omni_platform.device_typeand usesProfilerActivity.NPU(provided bytorch_npu) on NPU devices, falling back toProfilerActivity.CUDAotherwise.Fixes #1484
Test plan
export_chrome_traceshould produce a valid trace file.ProfilerActivity.CUDAis still used.cc @gcanlin