Open wangxihuiok opened 4 years ago
Yes, TASO preprocesses the new weights for the convolution when it is fused with a batchnorm. You can see the preprocessing steps when using TASO to optimize a ConvNet with batchnorm (e.g., ResNet).
Thanks for your interests in TASO.
for example, graph optimization can fuse the conv, batchnorm layer to a new conv layer, and calculate the new fused conv weight parameters, can taso calculate the new weight in the transformation after fusing conv and batchnorm?