How to use pyconv in other backbones (e.g. EfficientNet). Can I directly replace all the convolutional layers with PyConv, then load the imagenet pre-trained parameters? Or need to train the modified network from scratch on ImageNet and then finetune on the downstream task?
Hi, thanks for your amazing work.
How to use pyconv in other backbones (e.g. EfficientNet). Can I directly replace all the convolutional layers with PyConv, then load the imagenet pre-trained parameters? Or need to train the modified network from scratch on ImageNet and then finetune on the downstream task?