how to freeze syncbn?

[xLuge]

hi, when i finetune the model,i need to freeze the syncbn. how can i do that?

[ravitejarj]

how r u fine tuning the model

[zhanghang1989]

hi, when i finetune the model,i need to freeze the syncbn. how can i do that?

You may use the eval mode. model.eval()

[xLuge]

how r u fine tuning the model
i just fine tuning model like that https://hangzhang.org/PyTorch-Encoding/model_zoo/segmentation.html#pascal-voc-dataset
but according to the deeplab3, we should freeze bn after coco pretrain.