Tags: siliconflow/onediff
Tags
Integrate deep cache (#426) ## ComfyUI Node and Example Node name: ModuleDeepCacheSpeedup Example workflow: https://github.com/Oneflow-Inc/onediff/blob/integrate_deep_cache/onediff_comfy_nodes/workflows/deep-cache.png ## Depending on the new oneflow community edition cuda 11.8: `python3 -m pip install --pre oneflow -f https://oneflow-pro.oss-cn-beijing.aliyuncs.com/branch/community/cu118` cuda12.1: `python3 -m pip install --pre oneflow -f https://oneflow-pro.oss-cn-beijing.aliyuncs.com/branch/community/cu121` cuda12.2: `python3 -m pip install --pre oneflow -f https://oneflow-pro.oss-cn-beijing.aliyuncs.com/branch/community/cu122` ## SDXL UNet iter speed ### a100 - 1024 * 1024 |SDXL(Pytorch)| DeepCache | SDXL(with OneDiff) |DeepCache with OneDiff| | :-: | :-: | :-: | :-: | |7.79it/s(1x)| 16.92it/s(2.17x) | 10.11it/s(1.30x) | 23.58it/s(3.03x) | - 512 * 512 |SDXL(Pytorch)| DeepCache | SDXL(with OneDiff) |DeepCache with OneDiff| | :-: | :-: | :-: | :-: | |22.47it/s(1x)| 48.73it/(2.17x) | 28.67it/s(1.26x) | 54.79it/s(2.44x) | ### 3090 - 1024 * 1024 |SDXL(Pytorch)| DeepCache | SDXL(with OneDiff) |DeepCache with OneDiff| | :-: | :-: | :-: | :-: | |4.01it/s(1x)| 8.79it/s(2.19x) | 6.41it/s(1.6x) | 14.11it/s(3.52x) | - 512 * 512 |SDXL(Pytorch)| DeepCache | SDXL(with OneDiff) |DeepCache with OneDiff| | :-: | :-: | :-: | :-: | |13.88it/(1x)| 29.81it/s(2.15x) | 21.49it/s(1.55x) | 43.56it/s(3.14x) | --------- Co-authored-by: FengWen <[email protected]> Co-authored-by: FengWen <[email protected]> Co-authored-by: Xiaoyu Xu <[email protected]>