The following problem occurs when I train train_siammask_refine ValueRrror:loaded state dict has a different number of parameter groups

[MacBookYang]

why?? could you help me? the resume model is my train model

[nanowhiter]

why?? could you help me? the resume model is my train model

Since the optimizer after 50% global epochs is different than before. If you want to resume the training with the snapshot after 50% global epochs, remember to modify the number of start_epoch by adding --start-epoch.