luckmatter/model_gen.py [196:214]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        super(ModelConv, self).__init__()
        self.ks = ks
        self.ws_linear = nn.ModuleList()
        self.ws_bn = nn.ModuleList()
        self.bn_before_relu = bn_before_relu

        init_k, h, w = input_size
        last_k = init_k

        for k in ks:
            k *= multi
            self.ws_linear.append(nn.Conv2d(last_k, k, 3))
            if has_bn:
                self.ws_bn.append(nn.BatchNorm2d(k))
            last_k = k
            h -= 2
            w -= 2

        self.final_w = nn.Linear(last_k * h * w, d_output)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



student_specialization/model_gen.py [218:236]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        super(ModelConv, self).__init__()
        self.ks = ks
        self.ws_linear = nn.ModuleList()
        self.ws_bn = nn.ModuleList()
        self.bn_before_relu = bn_before_relu

        init_k, h, w = input_size
        last_k = init_k

        for k in ks:
            k *= multi
            self.ws_linear.append(nn.Conv2d(last_k, k, 3))
            if has_bn:
                self.ws_bn.append(nn.BatchNorm2d(k))
            last_k = k
            h -= 2
            w -= 2

        self.final_w = nn.Linear(last_k * h * w, d_output)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -