model = Net() # 加载权重 ckpt=torch.load(weights, map_location=device) state_dict=ckpt.state_dict() # state_dict 是一个字典 # state_dict.keys() # odict_keys(['0.model.0.conv.conv.weight', '0.model.0.conv.conv.bias', '0.model.1.conv.weight', ...]) # 权重取舍处理 state_...
它是KAIMIN我试图打印出的权重，我发现它是初始化的默认与统一Kaiming，和权重是在conv1.lin_l.weight...