Python torch.diag方法代码示例

# 需要导入模块: import torch [as 别名]
# 或者: from torch import diag [as 别名]
def fuse_conv_and_bn(conv, bn):
    # https://tehnokv.com/posts/fusing-batchnorm-and-conv/
    with torch.no_grad():
        # init
        fusedconv = torch.nn.Conv2d(conv.in_channels,
                                    conv.out_channels,
                                    kernel_size=conv.kernel_size,
                                    stride=conv.stride,
                                    padding=conv.padding,
                                    bias=True)

        # prepare filters
        w_conv = conv.weight.clone().view(conv.out_channels, -1)
        w_bn = torch.diag(bn.weight.div(torch.sqrt(bn.eps + bn.running_var)))
        fusedconv.weight.copy_(torch.mm(w_bn, w_conv).view(fusedconv.weight.size()))

        # prepare spatial bias
        if conv.bias is not None:
            b_conv = conv.bias
        else:
            b_conv = torch.zeros(conv.weight.size(0))
        b_bn = bn.bias - bn.weight.mul(bn.running_mean).div(torch.sqrt(bn.running_var + bn.eps))
        fusedconv.bias.copy_(b_conv + b_bn)

        return fusedconv 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import diag [as 别名]
def _mix_rbf_kernel(X, Y, sigma_list):
    assert(X.size(0) == Y.size(0))
    m = X.size(0)

    Z = torch.cat((X, Y), 0)
    ZZT = torch.mm(Z, Z.t())
    diag_ZZT = torch.diag(ZZT).unsqueeze(1)
    Z_norm_sqr = diag_ZZT.expand_as(ZZT)
    exponent = Z_norm_sqr - 2 * ZZT + Z_norm_sqr.t()

    K = 0.0
    for sigma in sigma_list:
        gamma = 1.0 / (2 * sigma**2)
        K += torch.exp(-gamma * exponent)

    return K[:m, :m], K[:m, m:], K[m:, m:], len(sigma_list) 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import diag [as 别名]
def construct_diag(x: torch.Tensor):
    """
    Constructs a diagonal matrix based on batched data. Solution found here:
    https://stackoverflow.com/questions/47372508/how-to-construct-a-3d-tensor-where-every-2d-sub-tensor-is-a-diagonal-matrix-in-p
    Do note that it only considers the last axis.
    :param x: The tensor
    """

    if x.dim() < 1:
        return x
    elif x.shape[-1] < 2:
        return x.unsqueeze(-1)
    elif x.dim() < 2:
        return torch.diag(x)

    b = torch.eye(x.size(-1), device=x.device)
    c = x.unsqueeze(-1).expand(*x.size(), x.size(-1))

    return c * b 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import diag [as 别名]
def test_UnscentedTransform2D(self):
        # ===== 2D model ===== #
        mat = torch.eye(2)
        scale = torch.diag(mat)

        norm = Normal(0., 1.)
        mvn = MultivariateNormal(torch.zeros(2), torch.eye(2))
        mvnlinear = AffineProcess((fmvn, g), (mat, scale), mvn, mvn)
        mvnoblinear = AffineObservations((fomvn, gomvn), (1.,), norm)

        mvnmodel = StateSpaceModel(mvnlinear, mvnoblinear)

        # ===== Perform unscented transform ===== #
        uft = UnscentedFilterTransform(mvnmodel)
        res = uft.initialize(3000)
        p = uft.predict(res)
        c = uft.correct(0., p)

        assert isinstance(c.x_dist(), MultivariateNormal) and c.x_dist().mean.shape == torch.Size([3000, 2]) 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import diag [as 别名]
def __init__(self,in_channel):
        super(InvConv,self).__init__()

        weight=np.random.randn(in_channel,in_channel)
        q,_=linalg.qr(weight)
        w_p,w_l,w_u=linalg.lu(q.astype(np.float32))
        w_s=np.diag(w_u)
        w_u=np.triu(w_u,1)
        u_mask=np.triu(np.ones_like(w_u),1)
        l_mask=u_mask.T

        self.register_buffer('w_p',torch.from_numpy(w_p))
        self.register_buffer('u_mask',torch.from_numpy(u_mask))
        self.register_buffer('l_mask',torch.from_numpy(l_mask))
        self.register_buffer('l_eye',torch.eye(l_mask.shape[0]))
        self.register_buffer('s_sign',torch.sign(torch.from_numpy(w_s)))
        self.w_l=torch.nn.Parameter(torch.from_numpy(w_l))
        self.w_s=torch.nn.Parameter(torch.log(1e-7+torch.abs(torch.from_numpy(w_s))))
        self.w_u=torch.nn.Parameter(torch.from_numpy(w_u))

        self.weight=None
        self.invweight=None

        return 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import diag [as 别名]
def forward(self, input):
        laplacian = input.exp() + self.eps
        output = input.clone()
        for b in range(input.size(0)):
            lap = laplacian[b].masked_fill(
                Variable(torch.eye(input.size(1)).cuda().ne(0)), 0)
            lap = -lap + torch.diag(lap.sum(0))
            # store roots on diagonal
            lap[0] = input[b].diag().exp()
            inv_laplacian = lap.inverse()

            factor = inv_laplacian.diag().unsqueeze(1)\
                                         .expand_as(input[b]).transpose(0, 1)
            term1 = input[b].exp().mul(factor).clone()
            term2 = input[b].exp().mul(inv_laplacian.transpose(0, 1)).clone()
            term1[:, 0] = 0
            term2[0] = 0
            output[b] = term1 - term2
            roots_output = input[b].diag().exp().mul(
                inv_laplacian.transpose(0, 1)[0])
            output[b] = output[b] + torch.diag(roots_output)
        return output 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import diag [as 别名]
def __expm__(self, matrix, symmetric):
        r"""Calculates matrix exponential.

        Args:
            matrix (Tensor): Matrix to take exponential of.
            symmetric (bool): Specifies whether the matrix is symmetric.

        :rtype: (:class:`Tensor`)
        """
        if symmetric:
            e, V = torch.symeig(matrix, eigenvectors=True)
            diff_mat = V @ torch.diag(e.exp()) @ V.t()
        else:
            diff_mat_np = expm(matrix.cpu().numpy())
            diff_mat = torch.Tensor(diff_mat_np).to(matrix.device)
        return diff_mat 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import diag [as 别名]
def regularizer_orth2(m):
    """
    # ----------------------------------------
    # Applies regularization to the training by performing the
    # orthogonalization technique described in the paper
    # This function is to be called by the torch.nn.Module.apply() method,
    # which applies svd_orthogonalization() to every layer of the model.
    # usage: net.apply(regularizer_orth2)
    # ----------------------------------------
    """
    classname = m.__class__.__name__
    if classname.find('Conv') != -1:
        w = m.weight.data.clone()
        c_out, c_in, f1, f2 = w.size()
        # dtype = m.weight.data.type()
        w = w.permute(2, 3, 1, 0).contiguous().view(f1*f2*c_in, c_out)
        u, s, v = torch.svd(w)
        s_mean = s.mean()
        s[s > 1.5*s_mean] = s[s > 1.5*s_mean] - 1e-4
        s[s < 0.5*s_mean] = s[s < 0.5*s_mean] + 1e-4
        w = torch.mm(torch.mm(u, torch.diag(s)), v.t())
        m.weight.data = w.view(f1, f2, c_in, c_out).permute(3, 2, 0, 1)  # .type(dtype)
    else:
        pass 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import diag [as 别名]
def get_loadings(self) -> np.ndarray:
        """Extract per-gene weights (for each Z, shape is genes by dim(Z)) in the linear decoder."""
        # This is BW, where B is diag(b) batch norm, W is weight matrix
        if self.use_batch_norm is True:
            w = self.decoder.factor_regressor.fc_layers[0][0].weight
            bn = self.decoder.factor_regressor.fc_layers[0][1]
            sigma = torch.sqrt(bn.running_var + bn.eps)
            gamma = bn.weight
            b = gamma / sigma
            bI = torch.diag(b)
            loadings = torch.matmul(bI, w)
        else:
            loadings = self.decoder.factor_regressor.fc_layers[0][0].weight
        loadings = loadings.detach().cpu().numpy()
        if self.n_batch > 1:
            loadings = loadings[:, : -self.n_batch]

        return loadings 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import diag [as 别名]
def _regularizer(self, mu_z, std_z):
        kld_z = self.z_prior_stdv.log() - std_z.log() + (std_z ** 2 + (mu_z.pow(2) - self.z_prior_mean)) /\
                (2 * self.z_prior_stdv.pow(2)) - 0.5
        regularizer_loss = kld_z.sum()

        regularizer_loss = self.beta * regularizer_loss

        cov_mu_z = self._get_covariance_mu_z(mu_z)

        if self.mode == "i":
            dipvae_regularizer_loss = self._get_dipvae_regularizer(cov_mu_z, self.lambda_offdiag, self.lambda_diag)

        elif self.mode == "ii":
            cov_z = cov_mu_z + torch.mean(torch.diag(std_z**2), dim=0)
            dipvae_regularizer_loss = self._get_dipvae_regularizer(cov_z, self.lambda_offdiag, self.lambda_diag)
        else:
            raise NotImplementedError("Unsupported dipvae mode.")

        return regularizer_loss + dipvae_regularizer_loss 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import diag [as 别名]
def __init__(self, num_inputs):
        super(LUInvertibleMM, self).__init__()
        self.W = torch.Tensor(num_inputs, num_inputs)
        nn.init.orthogonal_(self.W)
        self.L_mask = torch.tril(torch.ones(self.W.size()), -1)
        self.U_mask = self.L_mask.t().clone()

        P, L, U = sp.linalg.lu(self.W.numpy())
        self.P = torch.from_numpy(P)
        self.L = nn.Parameter(torch.from_numpy(L))
        self.U = nn.Parameter(torch.from_numpy(U))

        S = np.diag(U)
        sign_S = np.sign(S)
        log_S = np.log(abs(S))
        self.sign_S = torch.from_numpy(sign_S)
        self.log_S = nn.Parameter(torch.from_numpy(log_S))

        self.I = torch.eye(self.L.size(0)) 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import diag [as 别名]
def forward(self, inputs, cond_inputs=None, mode='direct'):
        if str(self.L_mask.device) != str(self.L.device):
            self.L_mask = self.L_mask.to(self.L.device)
            self.U_mask = self.U_mask.to(self.L.device)
            self.I = self.I.to(self.L.device)
            self.P = self.P.to(self.L.device)
            self.sign_S = self.sign_S.to(self.L.device)

        L = self.L * self.L_mask + self.I
        U = self.U * self.U_mask + torch.diag(
            self.sign_S * torch.exp(self.log_S))
        W = self.P @ L @ U

        if mode == 'direct':
            return inputs @ W, self.log_S.sum().unsqueeze(0).unsqueeze(
                0).repeat(inputs.size(0), 1)
        else:
            return inputs @ torch.inverse(
                W), -self.log_S.sum().unsqueeze(0).unsqueeze(0).repeat(
                    inputs.size(0), 1) 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import diag [as 别名]
def test_np():
    npr.seed(0)

    nx, nineq, neq = 4, 6, 7
    Q = npr.randn(nx, nx)
    G = npr.randn(nineq, nx)
    A = npr.randn(neq, nx)
    D = np.diag(npr.rand(nineq))

    K_ = np.bmat((
        (Q, np.zeros((nx, nineq)), G.T, A.T),
        (np.zeros((nineq, nx)), D, np.eye(nineq), np.zeros((nineq, neq))),
        (G, np.eye(nineq), np.zeros((nineq, nineq + neq))),
        (A, np.zeros((neq, nineq + nineq + neq)))
    ))

    K = block((
        (Q,   0, G.T, A.T),
        (0,   D, 'I',   0),
        (G, 'I',   0,   0),
        (A,   0,   0,   0)
    ))

    assert np.allclose(K_, K) 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import diag [as 别名]
def loss_l2(self, l2=0):
        """L2 loss centered around mu_init, scaled optionally per-source.

        In other words, diagonal Tikhonov regularization,
            ||D(\mu-\mu_{init})||_2^2
        where D is diagonal.

        Args:
            - l2: A float or np.array representing the per-source regularization
                strengths to use
        """
        if isinstance(l2, (int, float)):
            D = l2 * torch.eye(self.d)
        else:
            D = torch.diag(torch.from_numpy(l2))

        # Note that mu is a matrix and this is the *Frobenius norm*
        return torch.norm(D @ (self.mu - self.mu_init)) ** 2 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import diag [as 别名]
def _set_class_balance(self, class_balance, Y_dev):
        """Set a prior for the class balance

        In order of preference:
        1) Use user-provided class_balance
        2) Estimate balance from Y_dev
        3) Assume uniform class distribution
        """
        if class_balance is not None:
            self.p = np.array(class_balance)
        elif Y_dev is not None:
            class_counts = Counter(Y_dev)
            sorted_counts = np.array([v for k, v in sorted(class_counts.items())])
            self.p = sorted_counts / sum(sorted_counts)
        else:
            self.p = (1 / self.k) * np.ones(self.k)
        self.P = torch.diag(torch.from_numpy(self.p)).float()