can pretrained spark params fit into a regular res50 model?

[DietDietDiet]

I am using spark as pretrain task, and just wondering if a pretrained spark can be loaded into regular res50 for downstream task. Thanks @fangyixiao18

[fangyixiao18]

Yes, it can be loaded.
You can check this config https://github.com/open-mmlab/mmpretrain/blob/main/configs/spark/benchmarks/resnet50_8xb256-coslr-300e_in1k.py