EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

[Main][Feat]Set the Profiler parameters through environment variables consistent with vLLM (#2608)

Browse Source 
### What this PR does / why we need it?
Currently, when performing profiling in vLLM-Ascend, if you need to
obtain the Python call stack, you have to manually modify the code. The
code location is:
[worker_v1.py#L337](6c973361fc/vllm_ascend/worker/worker_v1.py (L337))
where you set with_stack to true.
Now, in vLLM, you can set whether to obtain the Python call stack
through an environment variable. The relevant PR is:
[#21803](https://github.com/vllm-project/vllm/pull/21803) and the
documentation is:
[profiling](https://docs.vllm.ai/en/latest/contributing/profiling.html?h=vllm_torch_profiler_with_stack#profile-with-pytorch-profiler)
This PR sets the profiler initialization parameters by using the same
environment variable as vLLM, eliminating the need for manual code
modification.

### Does this PR introduce _any_ user-facing change?
No

### How was this patch tested?
CI passed with new added/existing test.

- vLLM version: v0.10.1.1
- vLLM main:
0235103cbb

---------

Signed-off-by: zhanghaiwen <zhanghaiwen@cmss.chinamobile.com>
Co-authored-by: zhanghaiwen <zhanghaiwen@cmss.chinamobile.com>
This commit is contained in:
zhanghw0354
2025-09-03 10:58:08 +08:00
committed by GitHub
parent 93754d8061
commit eaeb2efb20
2 changed files with 9 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
tests/ut/worker/test_worker_v1.py
View File

@@ -452,11 +452,13 @@ class TestNPUWorker(TestBase):
mock_logger,
mock_envs_vllm,
):
"""Test _init_profiler method - profiler enabled case"""
"""Test _init_profiler method - profiler enabled case with stack and memory profiling enabled"""
from vllm_ascend.worker.worker_v1 import NPUWorker
# Set environment variables to enable profiler
mock_envs_vllm.VLLM_TORCH_PROFILER_DIR = "/path/to/traces"
mock_envs_vllm.VLLM_TORCH_PROFILER_WITH_STACK = True
mock_envs_vllm.VLLM_TORCH_PROFILER_WITH_PROFILE_MEMORY = True
# Set enum mocks
mock_export_type.Text = "Text"
@@ -516,8 +518,8 @@ class TestNPUWorker(TestBase):
# Verify profiler parameters
expected_activities = ["CPU", "NPU"]
self.assertEqual(profile_kwargs["activities"], expected_activities)
self.assertFalse(profile_kwargs["with_stack"])
self.assertFalse(profile_kwargs["profile_memory"])
self.assertTrue(profile_kwargs["with_stack"])
self.assertTrue(profile_kwargs["profile_memory"])
self.assertFalse(profile_kwargs["with_modules"])
self.assertEqual(profile_kwargs["experimental_config"],
mock_experimental_config_instance)

7
vllm_ascend/worker/worker_v1.py
View File

@@ -334,8 +334,9 @@ class NPUWorker(WorkerBase):
torch_npu.profiler.ProfilerActivity.CPU,
torch_npu.profiler.ProfilerActivity.NPU,
],
with_stack=False,
profile_memory=False,
with_stack=envs_vllm.VLLM_TORCH_PROFILER_WITH_STACK,
profile_memory=envs_vllm.\
VLLM_TORCH_PROFILER_WITH_PROFILE_MEMORY,
with_modules=False,
experimental_config=experimental_config,
on_trace_ready=torch_npu.profiler.tensorboard_trace_handler(
@@ -350,4 +351,4 @@ class NPUWorker(WorkerBase):
return self.model_runner.get_supported_tasks()
def take_draft_token_ids(self) -> Optional[DraftTokenIds]:
return self.model_runner.take_draft_token_ids()
return self.model_runner.take_draft_token_ids()