逐渐衰减损失函数的权重

Gradually decay the weight of loss function

我不确定问这个问题的地方是否合适,如果我需要删除 post,请随时告诉我。

我是 pyTorch 的新手,目前在我的项目中使用 CycleGAN(pyTorch 实现),我了解 cycleGAN 的大部分实现。

我阅读了名为“CycleGAN with better Cycles”的论文,我正在尝试应用论文中提到的修改。修改之一是循环一致性权重衰减,我不知道如何应用。

optimizer_G.zero_grad()

# Identity loss
loss_id_A = criterion_identity(G_BA(real_A), real_A)
loss_id_B = criterion_identity(G_AB(real_B), real_B)

loss_identity = (loss_id_A + loss_id_B) / 2

# GAN loss
fake_B = G_AB(real_A)
loss_GAN_AB = criterion_GAN(D_B(fake_B), valid)
fake_A = G_BA(real_B)
loss_GAN_BA = criterion_GAN(D_A(fake_A), valid)

loss_GAN = (loss_GAN_AB + loss_GAN_BA) / 2

# Cycle consistency loss
recov_A = G_BA(fake_B)
loss_cycle_A = criterion_cycle(recov_A, real_A)
recov_B = G_AB(fake_A)
loss_cycle_B = criterion_cycle(recov_B, real_B)

loss_cycle = (loss_cycle_A + loss_cycle_B) / 2

# Total loss
loss_G =    loss_GAN + 
            lambda_cyc * loss_cycle + #lambda_cyc is 10
            lambda_id * loss_identity #lambda_id is 0.5 * lambda_cyc

loss_G.backward()
optimizer_G.step()

我的问题是如何逐渐衰减循环一致性损失的权重?

如能在实施此修改方面提供任何帮助,我们将不胜感激。

本文摘自: Cycle consistency loss 有助于在早期阶段稳定训练,但在后期阶段成为逼真图像的障碍。我们建议 随着训练的进行 逐渐衰减循环一致性损失 λ 的权重。然而,我们仍然应该确保 λ 是 不会衰减到 0,这样生成器就不会变得不受约束而变得完全疯狂。

提前致谢。

下面是您可以使用的原型函数!

def loss (other params, decay params, initial_lambda, steps):
    # compute loss
    # compute cyclic loss
    # function that computes lambda given the steps
    cur_lambda  = compute_lambda(step, decay_params, initial_lamdba) 

    final_loss = loss + cur_lambda*cyclic_loss 
    return final_loss

compute_lambda 函数以 50 步从 10 线性衰减到 1e-5

def compute_lambda(step, decay_params):
    final_lambda = decay_params["final"]
    initial_lambda = decay_params["initial"]
    total_step = decay_params["total_step"]
    start_step = decay_params["start_step"]

    if (step < start_step+total_step and step>start_step):
        return initial_lambda + (step-start_step)*(final_lambda-initial_lambda)/total_step
    elif (step < start_step):
        return initial_lambda 
    else:
        return final_lambda
# Usage:
compute_lambda(i, {"final": 1e-5, "initial":10, "total_step":50, "start_step" : 50})