Omni: ComfyUI 运行SenseNova模型会OOM
Open
- Dominant language
- C++
- Stars
- 529
- Forks
- 80
- Avg merge
- 9h 7m
- Merged PRs (30d)
- 38
Description
custom node : https://github.com/OpenSenseNova/ComfyUI-SenseNova-U1
workflow : https://github.com/OpenSenseNova/ComfyUI-SenseNova-U1/blob/main/example_workflows/sensenova_image_editing.json
使用这个工作流进行图片编辑
在Linux和Windows两个版本上测试，单图编辑是OK；双图编辑会OOM，显存使用超过32455 MB，利用率99%以上；
使用N卡进行测试，双图编辑也OK，显存使用31300 MB；
vram_mode 选择 fast
Contributor guide
Research direction
Start by running the linked ComfyUI-SenseNova-U1 example_workflows/sensenova_image_editing.json with vram_mode set to fast, comparing single-image and dual-image editing on the reported hardware. Measure VRAM usage and reproduce the dual-image OOM; done requires identifying a change that prevents the failure without breaking the working single-image workflow.
Written by the indexing model from the issue text.
Assessment
- Domain
- machine-learning, performance
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 45/100