Python torch.reciprocal方法代码示例

# 需要导入模块: import torch [as 别名]
# 或者: from torch import reciprocal [as 别名]
def get_mrr(indices, targets):
    """
    Calculates the MRR score for the given predictions and targets
    Args:
        indices (Bxk): torch.LongTensor. top-k indices predicted by the model.
        targets (B): torch.LongTensor. actual target indices.

    Returns:
        mrr (float): the mrr score
    """

    tmp = targets.view(-1, 1)
    targets = tmp.expand_as(indices)
    hits = (targets == indices).nonzero()
    ranks = hits[:, -1] + 1
    ranks = ranks.float()
    rranks = torch.reciprocal(ranks)
    mrr = torch.sum(rranks).data / targets.size(0)
    return mrr.item() 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import reciprocal [as 别名]
def _get_discount(self, slate_size: int) -> Tensor:
        weights = DCGSlateMetric._weights
        if (
            weights is None
            or weights.shape[0] < slate_size
            or weights.device != self._device
        ):
            DCGSlateMetric._weights = torch.reciprocal(
                torch.log2(
                    torch.arange(
                        2, slate_size + 2, dtype=torch.double, device=self._device
                    )
                )
            )
        weights = DCGSlateMetric._weights
        assert weights is not None
        return weights[:slate_size] 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import reciprocal [as 别名]
def evaluateMetric(ranks, metric):
    ranks = ranks.data.numpy()
    if metric == 'r1':
        ranks = ranks.reshape(-1)
        return 100 * (ranks == 1).sum() / float(ranks.shape[0])
    if metric == 'r5':
        ranks = ranks.reshape(-1)
        return 100 * (ranks <= 5).sum() / float(ranks.shape[0])
    if metric == 'r10':
        # ranks = ranks.view(-1)
        ranks = ranks.reshape(-1)
        # return 100*torch.sum(ranks <= 10).data[0]/float(ranks.size(0))
        return 100 * (ranks <= 10).sum() / float(ranks.shape[0])
    if metric == 'mean':
        # ranks = ranks.view(-1).float()
        ranks = ranks.reshape(-1).astype(float)
        return ranks.mean()
    if metric == 'mrr':
        # ranks = ranks.view(-1).float()
        ranks = ranks.reshape(-1).astype(float)
        # return torch.reciprocal(ranks).mean().data[0]
        return (1 / ranks).mean() 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import reciprocal [as 别名]
def anticoupling_law(self, a, b):
        return torch.mul(a, torch.reciprocal(b)) 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import reciprocal [as 别名]
def init_alpha_and_beta(self, init_beta):
        """
        Initialize the alpha and beta of quantization function.
        init_data in numpy format.
        """
        # activations initialization (obtained offline)
        self.beta.data = torch.Tensor([init_beta]).cuda()
        self.alpha.data = torch.reciprocal(self.beta.data)
        self.alpha_beta_inited = True 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import reciprocal [as 别名]
def quantizeConvParams(self, T, alpha, beta, init, train_phase):
        """
        quantize the parameters in forward
        """
        T = (T > 2000)*2000 + (T <= 2000)*T
        for index in range(self.num_of_params):
            if init:
                beta[index].data = torch.Tensor([self.threshold / self.target_modules[index].data.abs().max()]).cuda()
                alpha[index].data = torch.reciprocal(beta[index].data)
            # scale w
            x = self.target_modules[index].data.mul(beta[index].data)
            
            y = self.forward(x, T, self.QW_biases[index], train=train_phase)
            #scale w^hat
            self.target_modules[index].data = y.mul(alpha[index].data) 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import reciprocal [as 别名]
def updateQuaGradWeight(self, T, alpha, beta, init):
        """
        Calculate the gradients of all the parameters.
        The gradients of model parameters are saved in the [Variable].grad.data.
        Args:
            T: the temperature, a single number. 
            alpha: the scale factor of the output, a list.
            beta: the scale factor of the input, a list. 
            init: a flag represents the first loading of the quantization function.
        Returns:
            alpha_grad: the gradient of alpha.
            beta_grad: the gradient of beta.
        """
        beta_grad = [0.0] * len(beta)
        alpha_grad = [0.0] * len(alpha)
        T = (T > 2000)*2000 + (T <= 2000)*T 
        for index in range(self.num_of_params):
            if init:
                beta[index].data = torch.Tensor([self.threshold / self.target_modules[index].data.abs().max()]).cuda()
                alpha[index].data = torch.reciprocal(beta[index].data)
            x = self.target_modules[index].data.mul(beta[index].data)

            # set T = 1 when train binary model
            y_grad = self.backward(x, 1, self.QW_biases[index]).mul(T)
            # set T = T when train the other quantization model
            #y_grad = self.backward(x, T, self.QW_biases[index]).mul(T)
            
        
            beta_grad[index] = y_grad.mul(self.target_modules[index].data).mul(alpha[index].data).\
                               mul(self.target_modules[index].grad.data).sum()
            alpha_grad[index] = self.forward(x, T, self.QW_biases[index]).\
                                mul(self.target_modules[index].grad.data).sum()

            self.target_modules[index].grad.data = y_grad.mul(beta[index].data).mul(alpha[index].data).\
                                                   mul(self.target_modules[index].grad.data)
        return alpha_grad, beta_grad 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import reciprocal [as 别名]
def log_normal_diag(x, mean, log_var, average=False, reduce=True, dim=None):
    log_norm = -0.5 * (log_var + (x - mean) * (x - mean) * log_var.exp().reciprocal())
    if reduce:
        if average:
            return torch.mean(log_norm, dim)
        else:
            return torch.sum(log_norm, dim)
    else:
        return log_norm 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import reciprocal [as 别名]
def log_normal_normalized(x, mean, log_var, average=False, reduce=True, dim=None):
    log_norm = -(x - mean) * (x - mean)
    log_norm *= torch.reciprocal(2.*log_var.exp())
    log_norm += -0.5 * log_var
    log_norm += -0.5 * torch.log(2. * PI)

    if reduce:
        if average:
            return torch.mean(log_norm, dim)
        else:
            return torch.sum(log_norm, dim)
    else:
        return log_norm 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import reciprocal [as 别名]
def log_normal_normalized(x, mean, log_var, average=False, reduce=True, dim=None):
    log_norm = -(x - mean) * (x - mean)
    log_norm *= torch.reciprocal(2. * log_var.exp())
    log_norm += -0.5 * log_var
    log_norm += -0.5 * torch.log(2. * PI)

    if reduce:
        if average:
            return torch.mean(log_norm, dim)
        else:
            return torch.sum(log_norm, dim)
    else:
        return log_norm 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import reciprocal [as 别名]
def aten_reciprocal(inputs, attributes, scope):
    inp = inputs[0]
    ctx = current_context()
    net = ctx.network
    if ctx.is_tensorrt and has_trt_tensor(inputs):
        layer = net.add_unary(inp, trt.UnaryOperation.RECIP)
        output = layer.get_output(0)
        output.name = scope
        layer.name = scope
        return [output]
    elif ctx.is_tvm and has_tvm_tensor(inputs):
        raise NotImplementedError

    return [torch.reciprocal(inp)] 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import reciprocal [as 别名]
def give_laplacian_coordinates(self, pred, block_id):
        r''' Returns the laplacian coordinates for the predictions and given block.

            The helper matrices are used to detect neighbouring vertices and
            the number of neighbours which are relevant for the weight matrix.
            The maximal number of neighbours is 8, and if a vertex has less,
            the index -1 is used which points to the added zero vertex.

        Arguments:
            pred (tensor): vertex predictions
            block_id (int): deformation block id (1,2 or 3)
        '''
        batch_size = pred.shape[0]
        num_vert = pred.shape[1]
        # Add "zero vertex" for vertices with less than 8 neighbours
        vertex = torch.cat(
            [pred, torch.zeros(batch_size, 1, 3).to(self.device)], 1)
        assert(vertex.shape == (batch_size, num_vert+1, 3))
        # Get 8 neighbours for each vertex; if a vertex has less, the
        # remaining indices are -1
        indices = torch.from_numpy(
            self.lape_idx[block_id-1][:, :8]).to(self.device)
        assert(indices.shape == (num_vert, 8))
        weights = torch.from_numpy(
            self.lape_idx[block_id-1][:, -1]).float().to(self.device)
        weights = torch.reciprocal(weights)
        weights = weights.view(-1, 1).expand(-1, 3)
        vertex_select = vertex[:, indices.long(), :]
        assert(vertex_select.shape == (batch_size, num_vert, 8, 3))
        laplace = vertex_select.sum(dim=2)  # Add neighbours
        laplace = torch.mul(laplace, weights)  # Multiply by weights
        laplace = torch.sub(pred, laplace)  # Subtract from prediction
        assert(laplace.shape == (batch_size, num_vert, 3))
        return laplace 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import reciprocal [as 别名]
def step_function(self, x, y):
        return torch.reciprocal(1 + torch.exp(-self.k * (x - y))) 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import reciprocal [as 别名]
def reciprocal(t):
    """
    Element-wise reciprocal computed using cross-approximation; see PyTorch's `reciprocal()`.

    :param t: input :class:`Tensor`

    :return: a :class:`Tensor`
    """

    return tn.cross(lambda x: torch.reciprocal(x), tensors=t, verbose=False) 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import reciprocal [as 别名]
def rsqrt(t):
    """
    Element-wise square-root reciprocal computed using cross-approximation; see PyTorch's `rsqrt()`.

    :param t: input :class:`Tensor`

    :return: a :class:`Tensor`
    """

    return tn.cross(lambda x: torch.rsqrt(x), tensors=t, verbose=False)