Queaten about SS for pytorch

[yan_liu]

sorry for this issue.
I can't find the SS_N implementation. Did you haven't realized this part yet?

[Yaoyao Liu]

Thanks for your interest in our work.

The implementation for SS operations is in models/conv2d_mtl.py file. E.g.,

meta-transfer-learning/pytorch/models/conv2d_mtl.py

Lines 93 to 101 in 0cb2c39

 def forward(self, inp): 
     new_mtl_weight = self.mtl_weight.expand(self.weight.shape) 
     new_weight = self.weight.mul(new_mtl_weight) 
     if self.bias is not None: 
         new_bias = self.bias + self.mtl_bias 
     else: 
         new_bias = None 
     return F.conv2d(inp, new_weight, new_bias, self.stride, 
                     self.padding, self.dilation, self.groups)