losses.py

import torch
import torch.nn as nn


def calc_jacard_sim(label_a, label_b):
    if label_a.shape != label_b.shape:
        raise ValueError('Shapes are not the same')
    if len(label_a.shape) > 1:
        dim=1
    else:
        dim=0
    stacked = torch.stack((label_a, label_b), dim=0)
    upper = torch.min(stacked, dim=0)[0].sum(dim=dim).float()
    lower = torch.max(stacked, dim=0)[0].sum(dim=dim).float()
    value = torch.div(upper, lower)
    return value

class MultiSupConLoss(nn.Module):
    """Supervised Contrastive Learning: https://arxiv.org/pdf/2004.11362.pdf.
    It also supports the unsupervised contrastive loss in SimCLR
    Author: Yonglong Tian (yonglong@mit.edu)
    Date: May 07, 2020"""
    def __init__(self, temperature=0.07, contrast_mode='all',
                 base_temperature=0.07, c_treshold=0.3):
        super(MultiSupConLoss, self).__init__()
        self.temperature = temperature
        self.contrast_mode = contrast_mode
        self.base_temperature = base_temperature
        self.c_treshold = c_treshold

    def forward(self, features, labels=None, mask=None, multi=True):
        """Compute loss for model. If both `labels` and `mask` are None,
        it degenerates to SimCLR unsupervised loss:
        https://arxiv.org/pdf/2002.05709.pdf
        Args:
            features: hidden vector of shape [bsz, n_views, ...].
            labels: ground truth of shape [bsz].
            mask: contrastive mask of shape [bsz, bsz], mask_{i,j}=1 if sample j
                has the same class as sample i. Can be asymmetric.
        Returns:
            A loss scalar.
        """
        device = (torch.device('cuda')
                  if features.is_cuda
                  else torch.device('cpu'))

        if len(features.shape) < 3:
            raise ValueError('`features` needs to be [bsz, n_views, ...],'
                             'at least 3 dimensions are required')
        if len(features.shape) > 3:
            features = features.view(features.shape[0], features.shape[1], -1)

        batch_size = features.shape[0]
        
        if labels is not None and mask is not None:
            raise ValueError('Cannot define both `labels` and `mask`')
        elif labels is None and mask is None:
            mask = torch.eye(batch_size, dtype=torch.float32).to(device)
            multi_val = torch.ones_like(mask).to(device)
        elif labels is not None:
            #labels = labels.contiguous().view(-1, 1)
            #if labels.shape[0] != batch_size:
            #    raise ValueError('Num of labels does not match num of features')
            #mask = torch.eq(labels, labels.T).float().to(device)
            if len(labels.shape) < 2:
                raise ValueError('This loss only works with multi-label problem')
            labels = labels.contiguous()
            if labels.shape[0] != batch_size:
                raise ValueError('Num of labels does not match num of features')
            multi_labels = torch.zeros((labels.shape[0], labels.shape[0]), dtype=torch.float32).to(device)
            for x in range(labels.shape[0]):
                for y in range(labels.shape[0]):
                    multi_labels[x,y] = calc_jacard_sim(labels[x], labels[y])
            mask = torch.where(multi_labels >= self.c_treshold, 1., 0.)
            if multi:
                multi_val = multi_labels
            else:
                multi_val = torch.ones_like(multi_labels).to(device)
        else:
            mask = mask.float().to(device)
            multi_val = torch.ones_like(mask).to(device)

        contrast_count = features.shape[1]
        contrast_feature = torch.cat(torch.unbind(features, dim=1), dim=0)
        if self.contrast_mode == 'one':
            anchor_feature = features[:, 0]
            anchor_count = 1
        elif self.contrast_mode == 'all':
            anchor_feature = contrast_feature
            anchor_count = contrast_count
        else:
            raise ValueError('Unknown mode: {}'.format(self.contrast_mode))

        # compute logits
        anchor_dot_contrast = torch.div(
            torch.matmul(anchor_feature, contrast_feature.T),
            self.temperature)
        # for numerical stability
        logits_max, _ = torch.max(anchor_dot_contrast, dim=1, keepdim=True)
        logits = anchor_dot_contrast - logits_max.detach()

        # tile mask and mult_val
        mask = mask.repeat(anchor_count, contrast_count)
        multi_val = multi_val.repeat(anchor_count, contrast_count)
        # mask-out self-contrast cases
        logits_mask = torch.scatter(
            torch.ones_like(mask),
            1,
            torch.arange(batch_size * anchor_count).view(-1, 1).to(device),
            0
        )
        
        mask = mask * logits_mask
        
        # compute log_prob
        exp_logits = torch.exp(logits) * logits_mask
        log_prob = logits - torch.log(exp_logits.sum(1, keepdim=True))

        # multiply by mutli-label parameter
        multi_log_prob = log_prob * multi_val
        
        # compute mean of log-likelihood over positive
        mean_multi_log_prob_pos = (mask * multi_log_prob).sum(1) / mask.sum(1)
        # loss
        loss = - (self.temperature / self.base_temperature) * mean_multi_log_prob_pos
        loss = loss.view(anchor_count, batch_size)

        return loss.mean()


class AsymmetricLoss(nn.Module):
    def __init__(self, gamma_neg=4, gamma_pos=1, clip=0.05, eps=1e-8, disable_torch_grad_focal_loss=True):
        super(AsymmetricLoss, self).__init__()

        self.gamma_neg = gamma_neg
        self.gamma_pos = gamma_pos
        self.clip = clip
        self.disable_torch_grad_focal_loss = disable_torch_grad_focal_loss
        self.eps = eps

    def forward(self, x, y):
        """"
        Parameters
        ----------
        x: input logits
        y: targets (multi-label binarized vector)
        """

        # Calculating Probabilities
        x_sigmoid = torch.sigmoid(x)
        xs_pos = x_sigmoid
        xs_neg = 1 - x_sigmoid

        # Asymmetric Clipping
        if self.clip is not None and self.clip > 0:
            xs_neg = (xs_neg + self.clip).clamp(max=1)

        # Basic CE calculation
        los_pos = y * torch.log(xs_pos.clamp(min=self.eps))
        los_neg = (1 - y) * torch.log(xs_neg.clamp(min=self.eps))
        loss = los_pos + los_neg

        # Asymmetric Focusing
        if self.gamma_neg > 0 or self.gamma_pos > 0:
            if self.disable_torch_grad_focal_loss:
                torch.set_grad_enabled(False)
            pt0 = xs_pos * y
            pt1 = xs_neg * (1 - y)  # pt = p if t > 0 else 1-p
            pt = pt0 + pt1
            one_sided_gamma = self.gamma_pos * y + self.gamma_neg * (1 - y)
            one_sided_w = torch.pow(1 - pt, one_sided_gamma)
            if self.disable_torch_grad_focal_loss:
                torch.set_grad_enabled(True)
            loss *= one_sided_w

        return -loss.mean()


class AsymmetricLossOptimized(nn.Module):
    ''' Notice - optimized version, minimizes memory allocation and gpu uploading,
    favors inplace operations'''

    def __init__(self, gamma_neg=4, gamma_pos=1, clip=0.05, eps=1e-8, disable_torch_grad_focal_loss=False):
        super(AsymmetricLossOptimized, self).__init__()

        self.gamma_neg = gamma_neg
        self.gamma_pos = gamma_pos
        self.clip = clip
        self.disable_torch_grad_focal_loss = disable_torch_grad_focal_loss
        self.eps = eps

        # prevent memory allocation and gpu uploading every iteration, and encourages inplace operations
        self.targets = self.anti_targets = self.xs_pos = self.xs_neg = self.asymmetric_w = self.loss = None

    def forward(self, x, y):
        """"
        Parameters
        ----------
        x: input logits
        y: targets (multi-label binarized vector)
        """

        self.targets = y
        self.anti_targets = 1 - y

        # Calculating Probabilities
        self.xs_pos = torch.sigmoid(x)
        self.xs_neg = 1.0 - self.xs_pos

        # Asymmetric Clipping
        if self.clip is not None and self.clip > 0:
            self.xs_neg.add_(self.clip).clamp_(max=1)

        # Basic CE calculation
        self.loss = self.targets * torch.log(self.xs_pos.clamp(min=self.eps))
        self.loss.add_(self.anti_targets * torch.log(self.xs_neg.clamp(min=self.eps)))

        # Asymmetric Focusing
        if self.gamma_neg > 0 or self.gamma_pos > 0:
            if self.disable_torch_grad_focal_loss:
                torch.set_grad_enabled(False)
            self.xs_pos = self.xs_pos * self.targets
            self.xs_neg = self.xs_neg * self.anti_targets
            self.asymmetric_w = torch.pow(1 - self.xs_pos - self.xs_neg,
                                          self.gamma_pos * self.targets + self.gamma_neg * self.anti_targets)
            if self.disable_torch_grad_focal_loss:
                torch.set_grad_enabled(True)
            self.loss *= self.asymmetric_w

        return -self.loss.sum()


class ASLSingleLabel(nn.Module):
    '''
    This loss is intended for single-label classification problems
    '''
    def __init__(self, gamma_pos=0, gamma_neg=4, eps: float = 0.1, reduction='mean'):
        super(ASLSingleLabel, self).__init__()

        self.eps = eps
        self.logsoftmax = nn.LogSoftmax(dim=-1)
        self.targets_classes = []
        self.gamma_pos = gamma_pos
        self.gamma_neg = gamma_neg
        self.reduction = reduction

    def forward(self, inputs, target):
        '''
        "input" dimensions: - (batch_size,number_classes)
        "target" dimensions: - (batch_size)
        '''
        num_classes = inputs.size()[-1]
        log_preds = self.logsoftmax(inputs)
        self.targets_classes = torch.zeros_like(inputs).scatter_(1, target.long().unsqueeze(1), 1)

        # ASL weights
        targets = self.targets_classes
        anti_targets = 1 - targets
        xs_pos = torch.exp(log_preds)
        xs_neg = 1 - xs_pos
        xs_pos = xs_pos * targets
        xs_neg = xs_neg * anti_targets
        asymmetric_w = torch.pow(1 - xs_pos - xs_neg,
                                 self.gamma_pos * targets + self.gamma_neg * anti_targets)
        log_preds = log_preds * asymmetric_w

        if self.eps > 0:  # label smoothing
            self.targets_classes = self.targets_classes.mul(1 - self.eps).add(self.eps / num_classes)

        # loss calculation
        loss = - self.targets_classes.mul(log_preds)

        loss = loss.sum(dim=-1)
        if self.reduction == 'mean':
            loss = loss.mean()

        return loss


if __name__ == "__main__":
    batch_size = 12
    views = 2
    classes = 10
    criterion = MultiSupConLoss()
    features = torch.nn.functional.normalize(torch.rand(batch_size, views, 128), p=2.0, dim=2)
    labels = torch.randint(0,2, (batch_size, classes))
    
    # MultiSupContrast
    loss = criterion(features, labels)
    print(loss)
    # SupContrast
    loss = criterion(features, labels, multi=False)
    print(loss)
    # or SimCLR
    loss = criterion(features, multi=False)
    print(loss)