98e790696877d8065842d83f1819ae9c989a8c98,memcnn/models/revop.py,ReversibleBlockFunction,backward,#Any#Any#,117

Before Change

        z1_stop = Variable(y1.data, requires_grad=True)

        G_z1 = Gm.forward(z1_stop)
        x2 = y2 - G_z1
        x2_stop = Variable(x2.data, requires_grad=True)

        F_x2 = Fm.forward(x2_stop)
        x1 = y1 - F_x2
        x1_stop = Variable(x1.data, requires_grad=True)

        // Compute outputs building a sub-graph
        z1 = x1_stop + F_x2
        y2_ = x2_stop + G_z1
        y1_ = z1

        // Perform full backward pass on graph...
        y = torch.cat([y1_, y2_], dim=1)
        dd = torch.autograd.grad(y, (x1_stop, x2_stop) + tuple(Gm.parameters()) + tuple(Fm.parameters()), grad_output, retain_graph=False)

After Change

        // compute outputs building a sub-graph
        x1_, x2_ = torch.chunk(x_stop, chunks=2, dim=1)
        y1_ = x1_ + Fm.forward(x2_)
        y2_ = x2_ + Gm.forward(y1_)
        y = torch.cat([y1_, y2_], dim=1)

        // perform full backward pass on graph...

Instances