focal_loss.py

import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.cuda.amp as amp


##
# version 1: use torch.autograd
class FocalLossV1(nn.Module):

    def __init__(self,
                 alpha=0.25,
                 gamma=2,
                 reduction='mean',):
        super(FocalLossV1, self).__init__()
        self.alpha = alpha
        self.gamma = gamma
        self.reduction = reduction
        self.crit = nn.BCEWithLogitsLoss(reduction='none')

    def forward(self, logits, label):
        '''
        Usage is same as nn.BCEWithLogits:
            >>> criteria = FocalLossV1()
            >>> logits = torch.randn(8, 19, 384, 384)
            >>> lbs = torch.randint(0, 2, (8, 19, 384, 384)).float()
            >>> loss = criteria(logits, lbs)
        '''
        probs = torch.sigmoid(logits)
        coeff = torch.abs(label - probs).pow(self.gamma).neg()
        log_probs = torch.where(logits >= 0,
                F.softplus(logits, -1, 50),
                logits - F.softplus(logits, 1, 50))
        log_1_probs = torch.where(logits >= 0,
                -logits + F.softplus(logits, -1, 50),
                -F.softplus(logits, 1, 50))
        loss = label * self.alpha * log_probs + (1. - label) * (1. - self.alpha) * log_1_probs
        loss = loss * coeff

        if self.reduction == 'mean':
            loss = loss.mean()
        if self.reduction == 'sum':
            loss = loss.sum()
        return loss


##
# version 2: user derived grad computation
class FocalSigmoidLossFuncV2(torch.autograd.Function):
    '''
    compute backward directly for better numeric stability
    '''
    @staticmethod
    @amp.custom_fwd(cast_inputs=torch.float32)
    def forward(ctx, logits, label, alpha, gamma):
        #  logits = logits.float()

        probs = torch.sigmoid(logits)
        coeff = (label - probs).abs_().pow_(gamma).neg_()
        log_probs = torch.where(logits >= 0,
                F.softplus(logits, -1, 50),
                logits - F.softplus(logits, 1, 50))
        log_1_probs = torch.where(logits >= 0,
                -logits + F.softplus(logits, -1, 50),
                -F.softplus(logits, 1, 50))
        ce_term1 = log_probs.mul_(label).mul_(alpha)
        ce_term2 = log_1_probs.mul_(1. - label).mul_(1. - alpha)
        ce = ce_term1.add_(ce_term2)
        loss = ce * coeff

        ctx.vars = (coeff, probs, ce, label, gamma, alpha)

        return loss

    @staticmethod
    @amp.custom_bwd
    def backward(ctx, grad_output):
        '''
        compute gradient of focal loss
        '''
        (coeff, probs, ce, label, gamma, alpha) = ctx.vars

        d_coeff = (label - probs).abs_().pow_(gamma - 1.).mul_(gamma)
        d_coeff.mul_(probs).mul_(1. - probs)
        d_coeff = torch.where(label < probs, d_coeff.neg(), d_coeff)
        term1 = d_coeff.mul_(ce)

        d_ce = label * alpha
        d_ce.sub_(probs.mul_((label * alpha).mul_(2).add_(1).sub_(label).sub_(alpha)))
        term2 = d_ce.mul(coeff)

        grads = term1.add_(term2)
        grads.mul_(grad_output)

        return grads, None, None, None


class FocalLossV2(nn.Module):

    def __init__(self,
                 alpha=0.25,
                 gamma=2,
                 reduction='mean'):
        super(FocalLossV2, self).__init__()
        self.alpha = alpha
        self.gamma = gamma
        self.reduction = reduction

    def forward(self, logits, label):
        '''
        Usage is same as nn.BCEWithLogits:
            >>> criteria = FocalLossV2()
            >>> logits = torch.randn(8, 19, 384, 384)
            >>> lbs = torch.randint(0, 2, (8, 19, 384, 384)).float()
            >>> loss = criteria(logits, lbs)
        '''
        loss = FocalSigmoidLossFuncV2.apply(logits, label, self.alpha, self.gamma)
        if self.reduction == 'mean':
            loss = loss.mean()
        if self.reduction == 'sum':
            loss = loss.sum()
        return loss


##
# version 3: implement wit cpp/cuda to save memory and accelerate
import focal_cpp # import torch before import cpp extension
class FocalSigmoidLossFuncV3(torch.autograd.Function):
    '''
    use cpp/cuda to accelerate and shrink memory usage
    '''
    @staticmethod
    @amp.custom_fwd(cast_inputs=torch.float32)
    def forward(ctx, logits, labels, alpha, gamma):
        #  logits = logits.float()
        loss = focal_cpp.focalloss_forward(logits, labels, gamma, alpha)
        ctx.variables = logits, labels, alpha, gamma
        return loss

    @staticmethod
    @amp.custom_bwd
    def backward(ctx, grad_output):
        '''
        compute gradient of focal loss
        '''
        logits, labels, alpha, gamma = ctx.variables
        grads = focal_cpp.focalloss_backward(grad_output, logits, labels, gamma, alpha)
        return grads, None, None, None


class FocalLossV3(nn.Module):
    '''
    This use better formula to compute the gradient, which has better numeric stability. Also use cuda to shrink memory usage and accelerate.
    '''
    def __init__(self,
                 alpha=0.25,
                 gamma=2,
                 reduction='mean'):
        super(FocalLossV3, self).__init__()
        self.alpha = alpha
        self.gamma = gamma
        self.reduction = reduction

    def forward(self, logits, label):
        '''
        Usage is same as nn.BCEWithLogits:
            >>> criteria = FocalLossV3()
            >>> logits = torch.randn(8, 19, 384, 384)
            >>> lbs = torch.randint(0, 2, (8, 19, 384, 384)).float()
            >>> loss = criteria(logits, lbs)
        '''
        loss = FocalSigmoidLossFuncV3.apply(logits, label, self.alpha, self.gamma)
        if self.reduction == 'mean':
            loss = loss.mean()
        if self.reduction == 'sum':
            loss = loss.sum()
        return loss





if __name__ == '__main__':
    import torchvision
    import torch
    import numpy as np
    import random
    torch.manual_seed(15)
    random.seed(15)
    np.random.seed(15)
    torch.backends.cudnn.deterministic = True

    class Model(nn.Module):
        def __init__(self):
            super(Model, self).__init__()
            net = torchvision.models.resnet18(pretrained=False)
            self.conv1 = net.conv1
            self.bn1 = net.bn1
            self.maxpool = net.maxpool
            self.relu = net.relu
            self.layer1 = net.layer1
            self.layer2 = net.layer2
            self.layer3 = net.layer3
            self.layer4 = net.layer4
            self.out = nn.Conv2d(512, 3, 3, 1, 1)
        def forward(self, x):
            feat = self.conv1(x)
            feat = self.bn1(feat)
            feat = self.relu(feat)
            feat = self.maxpool(feat)
            feat = self.layer1(feat)
            feat = self.layer2(feat)
            feat = self.layer3(feat)
            feat = self.layer4(feat)
            feat = self.out(feat)
            out = F.interpolate(feat, x.size()[2:], mode='bilinear', align_corners=True)
            return out
    net1 = Model()
    net2 = Model()
    net2.load_state_dict(net1.state_dict())

    criteria1 = FocalLossV2()
    criteria2 = FocalLossV3()
    net1.cuda()
    net2.cuda()
    net1.train()
    net2.train()
    net1.double()
    net2.double()
    criteria1.cuda()
    criteria2.cuda()

    optim1 = torch.optim.SGD(net1.parameters(), lr=1e-2)
    optim2 = torch.optim.SGD(net2.parameters(), lr=1e-2)

    bs = 16
    for it in range(300000):
        inten = torch.randn(bs, 3, 224, 244).cuda()
        #  lbs = torch.randint(0, 2, (bs, 3, 224, 244)).float().cuda()
        lbs = torch.randn(bs, 3, 224, 244).sigmoid().cuda()
        inten = inten.double()
        lbs = lbs.double()
        logits = net1(inten)
        loss1 = criteria1(logits, lbs)
        optim1.zero_grad()
        loss1.backward()
        optim1.step()
        logits = net2(inten)
        loss2 = criteria2(logits, lbs)
        optim2.zero_grad()
        loss2.backward()
        optim2.step()
        with torch.no_grad():
            if (it+1) % 50 == 0:
                print('iter: {}, ================='.format(it+1))
                print('out.weight: ', torch.mean(torch.abs(net1.out.weight - net2.out.weight)).item())
                print('conv1.weight: ', torch.mean(torch.abs(net1.conv1.weight - net2.conv1.weight)).item())
                print('loss: ', loss1.item() - loss2.item())