Dear Sir, hello. When reproducing your code, I would like to ask you the following questions
What is the torch version used in the experiment
According to the download from the readme, the SwinV2 ImageNet pre trained weights are in the swinv2 large version, while the code is in the swinv2 base version. Which version should be used for reproduction? Also, in the swinv2 large version, the model parameters cannot be matched. How can we solve this problem? As shown in the following figure
Dear Sir, hello. When reproducing your code, I would like to ask you the following questions