models/s2s_big_hier.py [211:256]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        self.det_init_connections = {
            0: 16,
            2: 13,
            4: 10,
            6: 7,
            8: 4,
            10: 1,
        }

        # Connection branches
        self.det_init_nets = nn.ModuleDict({
            'layer_16': nn.Sequential(
                layers.DcConv(n_hid*8*self.n_ctx, n_hid*8*self.n_ctx, 1),
                layers.TemporalConv2d(n_hid*8*self.n_ctx, n_hid*8*2, 1),
                layers.TemporalNorm2d(1, n_hid*8*2)
            ),
            'layer_13': nn.Sequential(
                layers.DcConv(n_hid*8*self.n_ctx, n_hid*8*self.n_ctx, 3, 1, 1),
                layers.TemporalConv2d(self.n_ctx*n_hid*8, 2*n_hid*8, 1),
                layers.TemporalNorm2d(16, n_hid*8*2)
            ),
            'layer_10': nn.Sequential(
                layers.DcConv(n_hid*8*self.n_ctx, n_hid*8*self.n_ctx, 1),
                layers.TemporalConv2d(n_hid*8*self.n_ctx, n_hid*8*2, 1),
                layers.TemporalNorm2d(16, n_hid*8*2)
            ),
            'layer_7': nn.Sequential(
                layers.DcConv(n_hid*4*self.n_ctx, n_hid*4*self.n_ctx, 1),
                layers.TemporalConv2d(n_hid*4*self.n_ctx, n_hid*4*2, 1),
                layers.TemporalNorm2d(16, n_hid*8)
            ),
            'layer_4': nn.Sequential(
                layers.DcConv(n_hid*2*self.n_ctx, n_hid*2*self.n_ctx, 1),
                layers.TemporalConv2d(n_hid*2*self.n_ctx, n_hid*2*2, 1),
                layers.TemporalNorm2d(16, n_hid*4)
            ),
            'layer_1': nn.Sequential(
                layers.DcConv(n_hid*1*self.n_ctx, n_hid*1*self.n_ctx, 1),
                layers.TemporalConv2d(n_hid*1*self.n_ctx, n_hid*1*2, 1),
                layers.TemporalNorm2d(16, n_hid*2)
            ),
        })

        # Stochastic connection list
        # encoder -> renderer
        self.sto_branches = {
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



models/s2s_big_hier_v4.py [236:281]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        self.det_init_connections = {
            0: 16,
            2: 13,
            4: 10,
            6: 7,
            8: 4,
            10: 1,
        }

        # Connection branches
        self.det_init_nets = nn.ModuleDict({
            'layer_16': nn.Sequential(
                layers.DcConv(n_hid*8*self.n_ctx, n_hid*8*self.n_ctx, 1),
                layers.TemporalConv2d(n_hid*8*self.n_ctx, n_hid*8*2, 1),
                layers.TemporalNorm2d(1, n_hid*8*2)
            ),
            'layer_13': nn.Sequential(
                layers.DcConv(n_hid*8*self.n_ctx, n_hid*8*self.n_ctx, 3, 1, 1),
                layers.TemporalConv2d(self.n_ctx*n_hid*8, 2*n_hid*8, 1),
                layers.TemporalNorm2d(16, n_hid*8*2)
            ),
            'layer_10': nn.Sequential(
                layers.DcConv(n_hid*8*self.n_ctx, n_hid*8*self.n_ctx, 1),
                layers.TemporalConv2d(n_hid*8*self.n_ctx, n_hid*8*2, 1),
                layers.TemporalNorm2d(16, n_hid*8*2)
            ),
            'layer_7': nn.Sequential(
                layers.DcConv(n_hid*4*self.n_ctx, n_hid*4*self.n_ctx, 1),
                layers.TemporalConv2d(n_hid*4*self.n_ctx, n_hid*4*2, 1),
                layers.TemporalNorm2d(16, n_hid*8)
            ),
            'layer_4': nn.Sequential(
                layers.DcConv(n_hid*2*self.n_ctx, n_hid*2*self.n_ctx, 1),
                layers.TemporalConv2d(n_hid*2*self.n_ctx, n_hid*2*2, 1),
                layers.TemporalNorm2d(16, n_hid*4)
            ),
            'layer_1': nn.Sequential(
                layers.DcConv(n_hid*1*self.n_ctx, n_hid*1*self.n_ctx, 1),
                layers.TemporalConv2d(n_hid*1*self.n_ctx, n_hid*1*2, 1),
                layers.TemporalNorm2d(16, n_hid*2)
            ),
        })

        # Stochastic connection list
        # encoder -> renderer
        self.sto_branches = {
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -