diffusers
[feat] implement `record_stream` when using CUDA streams during group offloading
#11081
Merged

Loading