4d6aeeca8a6397fb0c2ce1aa0679155c21c6cbbb,pytorch/pytorchcv/models/isqrtcovresnet.py,Triuvec,backward,#Any#Any#,128

Before Change

    @staticmethod
    def backward(ctx, grad_output):
        input, index = ctx.saved_tensors
        x = input
        batch_size = x.data.shape[0]
        channels = x.data.shape[1]
        grad_input = torch.zeros(batch_size, channels * channels, device=x.device, requires_grad=False).type(x.dtype)
        grad_input[:, index] = grad_output
        grad_input = grad_input.reshape(batch_size, channels, channels)
        return grad_input

After Change

    def backward(ctx, grad_y):
        x, triuvec_inds = ctx.saved_tensors
        batch, n, _ = x.size()
        grad_x = torch.zeros_like(x).view(batch, -1)
        grad_x[:, triuvec_inds] = grad_y
        grad_x = grad_x.view(batch, n, n)
        return grad_x

Instances