After inspection of the model parameters, I noticed that each BatchNorm layer inside a poly-3 module has a different set of parameters. For example parameters for layers inception_resnet_b_1_path1_1x7_bn, inception_resnet_b_1_poly2_block1_path1_1x7_bn and inception_resnet_b_1_poly2_block2_path1_1x7_bn differ while inception_resnet_b_1_path1_1x7, inception_resnet_b_1_poly2_block1_path1_1x7 and inception_resnet_b_1_poly2_block2_path1_1x7 Convolutional layers have the same parameters.
Is my understanding correct that respective Convolutional layers are shared between all Inception blocks inside a poly-3 module while each BatchNorm layer inside an Inception block has its own set of parameters and therefore parameters of BatchNorm layers are not shared?
After inspection of the model parameters, I noticed that each BatchNorm layer inside a poly-3 module has a different set of parameters. For example parameters for layers
inception_resnet_b_1_path1_1x7_bn,inception_resnet_b_1_poly2_block1_path1_1x7_bnandinception_resnet_b_1_poly2_block2_path1_1x7_bndiffer whileinception_resnet_b_1_path1_1x7,inception_resnet_b_1_poly2_block1_path1_1x7andinception_resnet_b_1_poly2_block2_path1_1x7Convolutional layers have the same parameters.Is my understanding correct that respective Convolutional layers are shared between all Inception blocks inside a poly-3 module while each BatchNorm layer inside an Inception block has its own set of parameters and therefore parameters of BatchNorm layers are not shared?