The following problem occurs when I train train_siammask_refine ValueRrror:loaded state dict has a different number of parameter groups
MacBookYang opened this issue · 1 comments
MacBookYang commented
why?? could you help me? the resume model is my train model
nanowhiter commented
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.