The EfficientNet is used to get feature embeddings. Are the EfficientNet and the transfomer trained at the same time?
If not, how to train the EfficientNet and what is the training label of EfficientNet?