chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[diffusion] profiling: simplify --perf-dump-path JSON output (remove duplicate denoise steps) (#15537)

Browse Source 
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
This commit is contained in:
Xiaoyu Zhang
2025-12-21 12:56:46 +08:00
committed by GitHub
parent 26704c23c0
commit 42bff706df
2 changed files with 5 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
python/sglang/multimodal_gen/docs/profiling.md
View File

@@ -67,6 +67,8 @@ Besides profiler traces, you can also dump a lightweight JSON report that contai
This is useful to quickly identify which stage dominates end-to-end latency, and whether denoising steps have uniform runtimes (and if not, which step has an abnormal spike).
The dumped JSON contains a `denoise_steps_ms` field formatted as an array of objects, each with a `step` key (the step index) and a `duration_ms` key.
Example:
```bash

8
python/sglang/multimodal_gen/runtime/utils/perf_logger.py
View File

@@ -207,10 +207,9 @@ class PerformanceLogger:
for name, duration_ms in timings.stages.items()
]
denoise_steps_ms = list(timings.steps)
denoise_steps = [
{"index": idx, "duration_ms": duration_ms}
for idx, duration_ms in enumerate(denoise_steps_ms)
denoise_steps_ms = [
{"step": idx, "duration_ms": duration_ms}
for idx, duration_ms in enumerate(timings.steps)
]
report = {
@@ -221,7 +220,6 @@ class PerformanceLogger:
"total_duration_ms": timings.total_duration_ms,
"steps": formatted_steps,
"denoise_steps_ms": denoise_steps_ms,
"denoise_steps": denoise_steps,
"meta": meta or {},
}