fe7a76d6创建于 2023年11月9日历史提交

文件	最后提交记录	最后更新时间
README.md	Handle enable CUDA graph param in SD example (#246) This PR updates how the enable_cuda_graph param is set depending on the world_size i.e. CUDA graphs should only be enabled when world_size==1.	3 年前
local_pipeline_stable_diffusion.py	Fix SD example imports for latest diffusers (#806)	2 年前
requirements.txt	Fix SD example imports for latest diffusers (#806)	2 年前
test-stable-diffusion.py	Update Stable Diffusion Example (#662) This PR updates the Stable Diffusion example with the following changes: Ability to switch between HF and local pipeline via use_local_pipe arg Added optional name arg to be able to specify model name from command line (default is prompthero/midjourney-v4-diffusion) Fixed local_rank issue by adding argument to script. Only enable CUDA graph when the local pipeline isn't used (not working w/ local pipeline at the moment)	3 年前

DeepSpeed Stable Diffusion Example

Setup

Python dependencies:

pip install -r requirements.txt

Examples can be run as follows:

deepspeed --num_gpus [number of GPUs] test-[model].py

NOTE: Local CUDA graphs for replaced SD modules will only be enabled when mp_size==1.

Command:

deepspeed --num_gpus 1 test-stable-diffusion.py

Output:

./baseline.png
./deepspeed.png