Python torch.ne方法代码示例

# 需要导入模块: import torch [as 别名]
# 或者: from torch import ne [as 别名]
def intersectionAndUnion(batch_data, pred, numClass):
    (imgs, segs, infos) = batch_data
    _, preds = torch.max(pred.data.cpu(), dim=1)

    # compute area intersection
    intersect = preds.clone()
    intersect[torch.ne(preds, segs)] = -1

    area_intersect = torch.histc(intersect.float(),
                                 bins=numClass,
                                 min=0,
                                 max=numClass - 1)

    # compute area union:
    preds[torch.lt(segs, 0)] = -1
    area_pred = torch.histc(preds.float(),
                            bins=numClass,
                            min=0,
                            max=numClass - 1)
    area_lab = torch.histc(segs.float(),
                           bins=numClass,
                           min=0,
                           max=numClass - 1)
    area_union = area_pred + area_lab - area_intersect
    return area_intersect, area_union 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import ne [as 别名]
def _augmented_scores(self, s, y):
        if self._range is None:
            delattr(self, '_range')
            self.register_buffer('_range', torch.arange(s.size(1), device=s.device)[None, :])

        delta = torch.ne(y[:, None], self._range).detach().float()
        return s + delta - s.gather(1, y[:, None]) 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import ne [as 别名]
def nNanElement(x):
    return torch.sum(torch.ne(x, x).float()) 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import ne [as 别名]
def getNanMask(x):
    return torch.ne(x, x) 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import ne [as 别名]
def forward(self, outputs, target, cout, epoch_num, inputs, *args):           
        val_pixels = torch.ne(target, 0).float().cuda()
        err = F.smooth_l1_loss(outputs*val_pixels, target*val_pixels, reduction='none')
        
        val_pixels = torch.ne(inputs, 0).float().cuda()
        inp_loss = F.smooth_l1_loss(outputs*val_pixels, inputs*val_pixels, reduction='none')
        
        loss = err + 0.1 * inp_loss
        
        return torch.mean(loss) 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import ne [as 别名]
def navg_forward(self, navg, c, x, b, eps=1e-20, restore=False):
                            
        # Normalized Averaging 
        ca = navg(c) 
        xout = torch.div(navg(x*c), ca + eps)
        
        # Add bias
        sz = b.size(0)
        b = b.view(1,sz,1,1)
        b = b.expand_as(xout)
        xout = xout + b
        
        if restore:
            cm = (c == 0).float()
            xout = torch.mul(xout, cm) + torch.mul(1-cm, x)
            
        # Propagate confidence
        #cout = torch.ne(ca, 0).float()
        cout = ca
        sz = cout.size()
        cout = cout.view(sz[0], sz[1], -1)
        
        k = navg.weight
        k_sz = k.size()
        k = k.view(k_sz[0], -1)
        s = torch.sum(k, dim=-1, keepdim=True)
        
        cout = cout / s 
        
        cout = cout.view(sz)
        k = k.view(k_sz)
        
        return xout, cout 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import ne [as 别名]
def process_data(self, batch):
        text1 = torch.LongTensor(batch['text1']).to(self.device)
        text2 = torch.LongTensor(batch['text2']).to(self.device)
        mask1 = torch.ne(text1, self.args.padding).unsqueeze(2)
        mask2 = torch.ne(text2, self.args.padding).unsqueeze(2)
        inputs = {
            'text1': text1,
            'text2': text2,
            'mask1': mask1,
            'mask2': mask2,
        }
        if 'target' in batch:
            target = torch.LongTensor(batch['target']).to(self.device)
            return inputs, target
        return inputs, None 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import ne [as 别名]
def pck(source_points,warped_points,L_pck,alpha=0.1):
    # compute precentage of correct keypoints
    batch_size=source_points.size(0)
    pck=torch.zeros((batch_size))
    for i in range(batch_size):
        p_src = source_points[i,:]
        p_wrp = warped_points[i,:]
        N_pts = torch.sum(torch.ne(p_src[0,:],-1)*torch.ne(p_src[1,:],-1))
        point_distance = torch.pow(torch.sum(torch.pow(p_src[:,:N_pts]-p_wrp[:,:N_pts],2),0),0.5)
        L_pck_mat = L_pck[i].expand_as(point_distance)
        correct_points = torch.le(point_distance,L_pck_mat*alpha)
        pck[i]=torch.mean(correct_points.float())
    return pck 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import ne [as 别名]
def mean_dist(source_points,warped_points,L_pck):
    # compute precentage of correct keypoints
    batch_size=source_points.size(0)
    dist=torch.zeros((batch_size))
    for i in range(batch_size):
        p_src = source_points[i,:]
        p_wrp = warped_points[i,:]
        N_pts = torch.sum(torch.ne(p_src[0,:],-1)*torch.ne(p_src[1,:],-1))
        point_distance = torch.pow(torch.sum(torch.pow(p_src[:,:N_pts]-p_wrp[:,:N_pts],2),0),0.5)
        L_pck_mat = L_pck[i].expand_as(point_distance)
        dist[i]=torch.mean(torch.div(point_distance,L_pck_mat))
    return dist 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import ne [as 别名]
def encode(self, features):
        seq = features["inputs"]
        pred = features["preds"]
        mask = torch.ne(seq, 0).float().cuda()
        enc_attn_bias = self.masking_bias(mask)

        inputs = torch.nn.functional.embedding(seq, self.embedding)

        if "embedding" in features and not self.training:
            embedding = features["embedding"]
            unk_mask = features["mask"].to(mask)[:, :, None]
            inputs = inputs * unk_mask + (1.0 - unk_mask) * embedding

        preds = torch.nn.functional.embedding(pred, self.weights)
        inputs = torch.cat([inputs, preds], axis=-1)
        inputs = inputs * (self.hidden_size ** 0.5)
        inputs = inputs + self.bias

        inputs = nn.functional.dropout(self.encoding(inputs), self.dropout,
                                       self.training)

        enc_attn_bias = enc_attn_bias.to(inputs)
        encoder_output = self.encoder(inputs, enc_attn_bias)
        logits = self.classifier(encoder_output)

        return logits 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import ne [as 别名]
def forward(self, features, labels):
        mask = torch.ne(features["inputs"], 0).float().cuda()
        logits = self.encode(features)
        loss = self.criterion(logits, labels)
        mask = mask.to(logits)

        return torch.sum(loss * mask) / torch.sum(mask) 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import ne [as 别名]
def delta(y, labels, alpha=None):
    """
    Compute zero-one loss matrix for a vector of ground truth y
    """

    if isinstance(y, ag.Variable):
        labels = ag.Variable(labels, requires_grad=False)

    delta = torch.ne(y[:, None], labels[None, :]).float()

    if alpha is not None:
        delta = alpha * delta
    return delta 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import ne [as 别名]
def split(x, y, labels):
    labels = ag.Variable(labels, requires_grad=False)
    mask = torch.ne(labels[None, :], y[:, None])

    # gather result:
    # x_1: all scores that do contain the ground truth
    x_1 = x[mask].view(x.size(0), -1)
    # x_2: scores of the ground truth
    x_2 = x.gather(1, y[:, None]).view(-1)
    return x_1, x_2 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import ne [as 别名]
def loss(self, x, sent_lengths, pos, rel, y):
        mask = torch.ne(x, self.pad_index)
        emissions = self.lstm_forward(x, pos, rel, sent_lengths)
        return self.crflayer(emissions, y, mask=mask) 

# 需要导入模块: import torch [as 别名]
# 或者: from torch import ne [as 别名]
def forward(self, x, poses, rels, sent_lengths):
        mask = torch.ne(x, self.pad_index)
        emissions = self.lstm_forward(x, poses, rels, sent_lengths)
        return self.crflayer.decode(emissions, mask=mask)