Python torch.trace方法代碼示例

# 需要導入模塊: import torch [as 別名]
# 或者: from torch import trace [as 別名]
def get_loss(pred, y, criterion, mtr, a=0.5):
    """
    To calculate loss
    :param pred: predicted value
    :param y: actual value
    :param criterion: nn.CrossEntropyLoss
    :param mtr: beta matrix
    """
    mtr_t = torch.transpose(mtr, 1, 2)
    aa = torch.bmm(mtr, mtr_t)
    loss_fn = 0
    for i in range(aa.size()[0]):
        aai = torch.add(aa[i, ], Variable(torch.neg(torch.eye(mtr.size()[1]))))
        loss_fn += torch.trace(torch.mul(aai, aai).data)
    loss_fn /= aa.size()[0]
    loss = torch.add(criterion(pred, y), Variable(torch.FloatTensor([loss_fn * a])))
    return loss 

# 需要導入模塊: import torch [as 別名]
# 或者: from torch import trace [as 別名]
def feature_smoothing(self, adj, X):
        adj = (adj.t() + adj)/2
        rowsum = adj.sum(1)
        r_inv = rowsum.flatten()
        D = torch.diag(r_inv)
        L = D - adj

        r_inv = r_inv  + 1e-3
        r_inv = r_inv.pow(-1/2).flatten()
        r_inv[torch.isinf(r_inv)] = 0.
        r_mat_inv = torch.diag(r_inv)
        # L = r_mat_inv @ L
        L = r_mat_inv @ L @ r_mat_inv

        XLXT = torch.matmul(torch.matmul(X.t(), L), X)
        loss_smooth_feat = torch.trace(XLXT)
        return loss_smooth_feat 

# 需要導入模塊: import torch [as 別名]
# 或者: from torch import trace [as 別名]
def batch_mat2angle(R):
    """ Calcuate the axis angles (twist) from a batch of rotation matrices

        Ethan Eade's lie group note:
        http://ethaneade.com/lie.pdf equation (17)

        function tested in 'test_geometry.py'

    :input
    :param Rotation matrix Bx3x3 \in \SO3 space
    --------
    :return 
    :param the axis angle B
    """
    R1 = [torch.trace(R[i]) for i in range(R.size()[0])]
    R_trace = torch.stack(R1)
    # clamp if the angle is too large (break small angle assumption)
    # @todo: not sure whether it is absoluately necessary in training. 
    angle = acos( ((R_trace - 1)/2).clamp(-1,1))
    return angle 

# 需要導入模塊: import torch [as 別名]
# 或者: from torch import trace [as 別名]
def evaluate(X_data, name='Eval'):
    model.eval()
    eval_idx_list = np.arange(len(X_data), dtype="int32")
    total_loss = 0.0
    count = 0
    with torch.no_grad():
        for idx in eval_idx_list:
            data_line = X_data[idx]
            x, y = Variable(data_line[:-1]), Variable(data_line[1:])
            if args.cuda:
                x, y = x.cuda(), y.cuda()
            output = model(x.unsqueeze(0)).squeeze(0)
            loss = -torch.trace(torch.matmul(y, torch.log(output).float().t()) +
                                torch.matmul((1-y), torch.log(1-output).float().t()))
            total_loss += loss.item()
            count += output.size(0)
        eval_loss = total_loss / count
        print(name + " loss: {:.5f}".format(eval_loss))
        return eval_loss 

# 需要導入模塊: import torch [as 別名]
# 或者: from torch import trace [as 別名]
def btrace(X):
    # batch-trace: [B, N, N] -> [B]
    n = X.size(-1)
    X_ = X.view(-1, n, n)
    tr = torch.zeros(X_.size(0)).to(X)
    for i in range(tr.size(0)):
        m = X_[i, :, :]
        tr[i] = torch.trace(m)
    return tr.view(*(X.size()[0:-2])) 

# 需要導入模塊: import torch [as 別名]
# 或者: from torch import trace [as 別名]
def torch_calculate_frechet_distance(mu1, sigma1, mu2, sigma2, eps=1e-6):
  """Pytorch implementation of the Frechet Distance.
  Taken from https://github.com/bioinf-jku/TTUR
  The Frechet distance between two multivariate Gaussians X_1 ~ N(mu_1, C_1)
  and X_2 ~ N(mu_2, C_2) is
          d^2 = ||mu_1 - mu_2||^2 + Tr(C_1 + C_2 - 2*sqrt(C_1*C_2)).
  Stable version by Dougal J. Sutherland.
  Params:
  -- mu1   : Numpy array containing the activations of a layer of the
             inception net (like returned by the function 'get_predictions')
             for generated samples.
  -- mu2   : The sample mean over activations, precalculated on an 
             representive data set.
  -- sigma1: The covariance matrix over activations for generated samples.
  -- sigma2: The covariance matrix over activations, precalculated on an 
             representive data set.
  Returns:
  --   : The Frechet Distance.
  """


  assert mu1.shape == mu2.shape, \
    'Training and test mean vectors have different lengths'
  assert sigma1.shape == sigma2.shape, \
    'Training and test covariances have different dimensions'

  diff = mu1 - mu2
  # Run 50 itrs of newton-schulz to get the matrix sqrt of sigma1 dot sigma2
  covmean = sqrt_newton_schulz(sigma1.mm(sigma2).unsqueeze(0), 50).squeeze()  
  out = (diff.dot(diff) +  torch.trace(sigma1) + torch.trace(sigma2)
         - 2 * torch.trace(covmean))
  return out


# Calculate Inception Score mean + std given softmax'd logits and number of splits 

# 需要導入模塊: import torch [as 別名]
# 或者: from torch import trace [as 別名]
def batch_mat2twist(R):
    """ The log map from SO3 to so3

        Calculate the twist vector from Rotation matrix 

        Ethan Eade's lie group note:
        http://ethaneade.com/lie.pdf equation (18)

        @todo: may rename the interface to batch_so3logmap(R)

        function tested in 'test_geometry.py'

        @note: it currently does not consider extreme small values. 
        If you use it as training loss, you may run into problems

    :input
    :param Rotation matrix Bx3x3 \in \SO3 space 
    --------
    :param the twist vector Bx3 \in \so3 space
    """
    B = R.size()[0]
 
    R1 = [torch.trace(R[i]) for i in range(R.size()[0])]
    tr = torch.stack(R1)
    theta = acos( ((tr - 1)/2).clamp(-1,1) )

    r11,r12,r13,r21,r22,r23,r31,r32,r33 = torch.split(R.view(B,-1),1,dim=1)
    res = torch.cat([r32-r23, r13-r31, r21-r12],dim=1)  

    magnitude = (0.5*theta/sin(theta))

    return magnitude.view(B,1) * res 

# 需要導入模塊: import torch [as 別名]
# 或者: from torch import trace [as 別名]
def forward(self):

        constrainted_matrix = self.select_param()
        matrix_ = torch.squeeze(torch.squeeze(constrainted_matrix,dim=2),dim=2)
        matrix_t = torch.t(matrix_)
        matrixs = torch.mm(matrix_t,matrix_)
        trace_ = torch.trace(torch.mm(matrixs,torch.inverse(matrixs)))
        log_det = torch.logdet(matrixs)
        maha_loss = trace_ - log_det
        return maha_loss 

# 需要導入模塊: import torch [as 別名]
# 或者: from torch import trace [as 別名]
def trace(self, a):
        return torch.trace(a) 

# 需要導入模塊: import torch [as 別名]
# 或者: from torch import trace [as 別名]
def block_trace(self, X, m, n):
        blocks = []
        for i in range(n):
            blocks.append([])
            for j in range(n):
                block = self.trace(X[..., i*m:(i+1)*m, j*m:(j+1)*m])
                blocks[-1].append(block)
        return self.pack([
            self.pack([
                b for b in block
            ])
            for block in blocks
        ]) 

# 需要導入模塊: import torch [as 別名]
# 或者: from torch import trace [as 別名]
def normalize_factors(A, B):
    eps = 1e-10

    trA = torch.trace(A) + eps
    trB = torch.trace(B) + eps
    assert trA > 0, 'Must PD. A not PD'
    assert trB > 0, 'Must PD. B not PD'
    return A * (trB/trA)**0.5, B * (trA/trB)**0.5 

# 需要導入模塊: import torch [as 別名]
# 或者: from torch import trace [as 別名]
def trace(tensor: BKTensor) -> BKTensor:
    new_real = torch.trace(tensor[0])
    new_imag = torch.trace(tensor[1])
    return torch.stack((new_real, new_imag)) 

# 需要導入模塊: import torch [as 別名]
# 或者: from torch import trace [as 別名]
def forward(self, ypred, ytrue):
		# geodesic loss between predicted and gt rotations
		tmp = torch.stack([torch.trace(torch.mm(ypred[i].t(), ytrue[i])) for i in range(ytrue.size(0))])
		angle = torch.acos(torch.clamp((tmp - 1.0) / 2, -1 + eps, 1 - eps))
		return torch.mean(angle) 

# 需要導入模塊: import torch [as 別名]
# 或者: from torch import trace [as 別名]
def my_loss(self, ypred, ytrue):
		# geodesic loss between predicted and gt rotations
		tmp = torch.stack([torch.trace(torch.mm(ypred[i].t(), ytrue[i])) for i in range(ytrue.size(0))])
		angle = torch.acos(torch.clamp((tmp - 1.0) / 2, -1 + eps, 1 - eps))
		return torch.mean(angle) 

# 需要導入模塊: import torch [as 別名]
# 或者: from torch import trace [as 別名]
def forward(self, adj_mat, l_mat):
        t0 = F.leaky_relu(self.encode0(adj_mat))
        t0 = F.leaky_relu(self.encode1(t0))
        self.embedding = t0
        t0 = F.leaky_relu(self.decode0(t0))
        t0 = F.leaky_relu(self.decode1(t0))
         
        L_1st = 2 * torch.trace(torch.mm(torch.mm(torch.t(self.embedding), l_mat), self.embedding))
        L_2nd = torch.sum(((adj_mat - t0) * adj_mat * self.beta) * ((adj_mat - t0) *  adj_mat * self.beta))
        
        L_reg = 0
        for param in self.parameters():
            L_reg += self.nu1 * torch.sum(torch.abs(param)) + self.nu2 * torch.sum(param * param)
        return self.alpha * L_1st,  L_2nd,  self.alpha * L_1st + L_2nd, L_reg