-
I am training FasterRCNN on my self-built dataset, and I already know that max_epochs are mentioned in train_cfg = dict(type='EpochBasedTrainLoop', max_epochs=12, val_interval=1). But I do have a question that how can I set a best max_epochs number, because I didn't find where there is a function like early_stop(If loss don't decrease in several epochs, then training will stop aotomatically.) |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
Hi, EarlyStopping feature is supported in EarlyStoppingHook. More details at https://mmengine.readthedocs.io/en/latest/api/generated/mmengine.hooks.EarlyStoppingHook.html#mmengine.hooks.EarlyStoppingHook |
Beta Was this translation helpful? Give feedback.
Hi, EarlyStopping feature is supported in EarlyStoppingHook. More details at https://mmengine.readthedocs.io/en/latest/api/generated/mmengine.hooks.EarlyStoppingHook.html#mmengine.hooks.EarlyStoppingHook