in the DA-transunet paper the number of trainable parameters for Transunet and DA-transunet are 105million and 107millions respectively. But when i write the model in tensorflow the number of trainable parameter of Transunet with resnet50 and without resnet wth the Image_size=(224,224,3), embed_dim=512, MLP size= 3072, num_head=12, num_transformer_layer=12 is 409million and 406 million. so i want to know where should i modify my code.
in the DA-transunet paper the number of trainable parameters for Transunet and DA-transunet are 105million and 107millions respectively. But when i write the model in tensorflow the number of trainable parameter of Transunet with resnet50 and without resnet wth the Image_size=(224,224,3), embed_dim=512, MLP size= 3072, num_head=12, num_transformer_layer=12 is 409million and 406 million. so i want to know where should i modify my code.