EfficientNet-PyTorch

EfficientNet-PyTorch copied to clipboard

Hello, I'm using efficientnet well. When I use this model, the purpose is to extract the activation layer value of the feature extractor part (. In other words, can the result of the activation layer output during training be changed by entering the code set_swish(memory_efficient=False), which requires a fixed weight at the time of training and inference?

thank you