factory/loss.py

import numpy as np

import torch
import torch.nn as nn
from torch.nn import functional as F

try:
    import torch.distributed.nn
    from torch import distributed as dist
    has_distributed = True
except ImportError:
    has_distributed = False

try:
    import horovod.torch as hvd
except ImportError:
    hvd = None


def gather_features(
        image_features,
        text_features,
        local_loss=False,
        gather_with_grad=False,
        rank=0,
        world_size=1,
        use_horovod=False
):
    assert has_distributed, 'torch.distributed did not import correctly, please use a PyTorch version with support.'
    if use_horovod:
        assert hvd is not None, 'Please install horovod'
        if gather_with_grad:
            all_image_features = hvd.allgather(image_features)
            all_text_features = hvd.allgather(text_features)
        else:
            with torch.no_grad():
                all_image_features = hvd.allgather(image_features)
                all_text_features = hvd.allgather(text_features)
            if not local_loss:
                # ensure grads for local rank when all_* features don't have a gradient
                gathered_image_features = list(all_image_features.chunk(world_size, dim=0))
                gathered_text_features = list(all_text_features.chunk(world_size, dim=0))
                gathered_image_features[rank] = image_features
                gathered_text_features[rank] = text_features
                all_image_features = torch.cat(gathered_image_features, dim=0)
                all_text_features = torch.cat(gathered_text_features, dim=0)
    else:
        # We gather tensors from all gpus
        if gather_with_grad:
            all_image_features = torch.cat(torch.distributed.nn.all_gather(image_features), dim=0)
            all_text_features = torch.cat(torch.distributed.nn.all_gather(text_features), dim=0)
        else:
            gathered_image_features = [torch.zeros_like(image_features) for _ in range(world_size)]
            gathered_text_features = [torch.zeros_like(text_features) for _ in range(world_size)]
            dist.all_gather(gathered_image_features, image_features)
            dist.all_gather(gathered_text_features, text_features)
            if not local_loss:
                # ensure grads for local rank when all_* features don't have a gradient
                gathered_image_features[rank] = image_features
                gathered_text_features[rank] = text_features
            all_image_features = torch.cat(gathered_image_features, dim=0)
            all_text_features = torch.cat(gathered_text_features, dim=0)

    return all_image_features, all_text_features


class ClipLoss(nn.Module):
    def __init__(
            self,
            local_loss=False,
            gather_with_grad=False,
            cache_labels=False,
            rank=0,
            world_size=1,
            use_horovod=False,
    ):
        super().__init__()
        self.local_loss = local_loss
        self.gather_with_grad = gather_with_grad
        self.cache_labels = cache_labels
        self.rank = rank
        self.world_size = world_size
        self.use_horovod = use_horovod

        # cache state
        self.prev_num_logits = 0
        self.labels = {}
        
    def forward(self, image_features, text_features):
        logit_scale = nn.Parameter(torch.ones([]) * np.log(1 / 0.07))
        device = image_features.device
        if self.world_size > 1:
            all_image_features, all_text_features = gather_features(
                image_features, text_features,
                self.local_loss, self.gather_with_grad, self.rank, self.world_size, self.use_horovod)

            if self.local_loss:
                logits_per_image = logit_scale * image_features @ all_text_features.T
                logits_per_text = logit_scale * text_features @ all_image_features.T
            else:
                logits_per_image = logit_scale * all_image_features @ all_text_features.T
                logits_per_text = logits_per_image.T
        else:
            logits_per_image = logit_scale * image_features @ text_features.T
            logits_per_text = logit_scale * text_features @ image_features.T

        # calculated ground-truth and cache if enabled
        num_logits = logits_per_image.shape[0]
        labels = torch.eye(num_logits, device=device, dtype=torch.float)
        pred_1 = F.log_softmax(logits_per_image,dim=-1)
        pred_2 = F.log_softmax(logits_per_text,dim=-1)
        loss_a = F.kl_div(pred_1, labels,reduction = 'sum')/num_logits
        loss_b = F.kl_div(pred_2, labels,reduction = 'sum')/num_logits
        total_loss = (loss_a + loss_b)/2
        return total_loss
    
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, use_weight = False):
        """"
        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
        if use_weight:
            return loss
        return -loss.sum()
    
class RalSingleLoss(nn.Module):
    '''
    This loss is intended for single-label classification problems
    '''
    def __init__(self, gamma_pos=0, gamma_neg=4, eps: float = 0.1, epsilon_pos_pow = -2.5, reduction='mean'):
        super(RalSingleLoss, 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
        self.epsilon_pos = 1.0
        self.epsilon_neg = 0.0
        self.epsilon_pos_pow = epsilon_pos_pow
        self.lamb = 1.5

    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 = torch.exp(log_preds)* (torch.log(xs_pos.clamp(min=self.eps)) + self.epsilon_pos * (1 - xs_pos.clamp(min=self.eps)) + self.epsilon_pos_pow * 0.5 * torch.pow(1 - xs_pos.clamp(min=self.eps), 2) ) * torch.log(xs_pos)
        xs_neg = (1 - xs_pos) * (torch.log(xs_neg.clamp(min=self.eps)) + self.epsilon_neg * (xs_neg.clamp(min=self.eps)) ) * -(self.lamb - xs_neg) * xs_neg ** 2
        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

class Ralloss(nn.Module):
    def __init__(self, gamma_neg=4, gamma_pos=0, clip=0.05, eps=1e-8, lamb=1.5, epsilon_neg=0.0, epsilon_pos=1.0, epsilon_pos_pow=-2.5, disable_torch_grad_focal_loss=False):
        super(Ralloss, 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

        # parameters of Taylor expansion polynomials
        self.epsilon_pos = epsilon_pos
        self.epsilon_neg = epsilon_neg
        self.epsilon_pos_pow = epsilon_pos_pow
        self.margin = 1.0
        self.lamb = lamb

    def forward(self, x, y, use_weight=False):
        """"
        x: input logits with size (batch_size, number of labels).
        y: binarized multi-label targets with size (batch_size, number of labels).
        """
        # 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 Taylor expansion polynomials
        los_pos = y * (torch.log(xs_pos.clamp(min=self.eps)) + self.epsilon_pos * (1 - xs_pos.clamp(min=self.eps)) + self.epsilon_pos_pow * 0.5 * torch.pow(1 - xs_pos.clamp(min=self.eps), 2))
        los_neg = (1 - y) * (torch.log(xs_neg.clamp(min=self.eps)) + self.epsilon_neg * (xs_neg.clamp(min=self.eps)) ) * (self.lamb - x_sigmoid) * x_sigmoid ** 2 * (self.lamb - xs_neg)
        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
        if use_weight:
            return loss
        return -loss.sum()