wangbo/ComfyUI
1
0
Fork 0
mirror of https://github.com/comfyanonymous/ComfyUI.git synced 2026-05-13 02:33:02 +08:00
Code Issues Actions 17 Packages Projects Releases Wiki Activity

Fix comments

Browse Source
This commit is contained in:
Tara Ding 2026-04-27 14:22:27 -07:00
parent d407d82350
commit 125ed0be4b
1 changed files with 9 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
benchmarks/benchmark_comfyui_serving.py
View File

@ -6,6 +6,15 @@ Submits prompts concurrently to a running ComfyUI server and reports
latency/throughput metrics. Input images and prompt files are prepared latency/throughput metrics. Input images and prompt files are prepared
automatically (and cached for reuse) before the benchmark starts. automatically (and cached for reuse) before the benchmark starts.
On first run the script will:
1. Download model weights (if --download-models is set).
2. Download the VBench I2V image dataset (requires: pip install gdown),
or generate synthetic placeholder images as a fallback.
3. Write one prompt JSON per input image under benchmarks/prompts/<model>_<task>/.
On subsequent runs all three steps are skipped if the files already exist.
Requests are distributed across prompt files in round-robin order.
Supported models / tasks Supported models / tasks
------------------------ ------------------------
wan22 / i2v — Wan 2.2 Image-to-Video (LightX2V 4-step, 720×720, 81 frames) wan22 / i2v — Wan 2.2 Image-to-Video (LightX2V 4-step, 720×720, 81 frames)