[Bugfix]: modify diffusion pipeline profiler result in videos

[bjf-frz]

PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS (AT THE BOTTOM) HAVE BEEN CONSIDERED.

Purpose

This PR addresses an issue where, after enabling the --enable-diffusion-pipeline-profiler, the /v1/videos interface in wan2.2 does not properly handle peak_memory and stage_durations.

Test Plan

server end:
vllm serve Wan2.2-I2V-A14B-Diffusers/ --omni --port 8091

user end:
python3 benchmarks/diffusion/diffusion_benchmark_serving.py
--base-url http://localhost:8091
--model Wan2.2-I2V-A14B-Diffusers/
--backend v1/videos
--dataset random
--task i2v
--num-prompts 1
--max-concurrency 1
--request-rate inf
--width 640
--height 480
--num-frames 81
--fps 16
--num-inference-steps 2

Test Result

================= Serving Benchmark Result =================
Backend:                                 v1/videos      
Model:                                   /home/admin/Wan2.2-I2V-14B-Distill-Diffusers/
Dataset:                                 random         
Task:                                    i2v            
--------------------------------------------------
Benchmark duration (s):                  12.14          
Request rate:                            inf            
Max request concurrency:                 1              
Successful requests:                     1/1              
--------------------------------------------------
Request throughput (req/s):              0.08           
Latency Mean (s):                        12.1411        
Latency Median (s):                      12.1411        
Latency P99 (s):                         12.1411        
Latency P95 (s):                         12.1411        
--------------------------------------------------
Peak Memory Max (MB):                    74204.00       
Peak Memory Mean (MB):                   74204.00       
Peak Memory Median (MB):                 74204.00       
--------------------------------------------------
Stage Durations Mean (s):
  Wan22I2VPipeline.text_encoder.forward: 0.0515         
  Wan22I2VPipeline.vae.encode:           1.0445         
  Wan22I2VPipeline.vae.decode:           1.5887

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan. Please provide the test scripts & test commands. Please state the reasons if your codes don't require additional test scripts. For test file guidelines, please check the test style doc
• The test results. Please paste the results comparison before and after, or the e2e results.
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model. Please run mkdocs serve to sync the documentation editions to ./docs.
• (Optional) Release notes update. If your change is user-facing, please update the release notes draft.

BEFORE SUBMITTING, PLEASE READ https://github.com/vllm-project/vllm-omni/blob/main/CONTRIBUTING.md (anything written below this line will be removed by GitHub Actions)

[chatgpt-codex-connector]

Codex usage limits have been reached for code reviews. Please check with the admins of this repo to increase the limits by adding credits.
Credits must be used to enable repository wide code reviews.

[david6666666]

please add purpose

[bjf-frz]

@Bounty-hunter @david6666666 @wtomin PTAL, thx !

[bjf-frz]

@yangjianjuan PTAL, thx

[david6666666]

One non-blocking test-coverage note below.

[david6666666]

Stage Durations Mean (s):
Wan22I2VPipeline.text_encoder.forward: 0.0515
Wan22I2VPipeline.vae.encode: 1.0445
Wan22I2VPipeline.vae.decode: 1.5887

Do we have dit Stage Durations time?

[bjf-frz]

@hsliuustc0106

Stage Durations Mean (s): Wan22I2VPipeline.text_encoder.forward: 0.0515 Wan22I2VPipeline.vae.encode: 1.0445 Wan22I2VPipeline.vae.decode: 1.5887

Do we have dit Stage Durations time?

The DIT process in wan2.2 is currently scattered throughout the forward. It needs to be refactored into a dedicated diffuse function to enable proper profiling. This refactoring will be addressed in a separate PR.

[david6666666]

LGTM