The weight of L1

[chunyu-li]

The loss function is $L = \lambda L_1 + L_2$. Could you please tell me the value of $\lambda$?

[alexLIUMinhao]

Hi,

This weight is used to balance the value of the two losses during the training process, so you can adjust it according to the actual values. The released model uses a value of 2.

[chunyu-li]

Hi,

This weight is used to balance the value of the two losses during the training process, so you can adjust it according to the actual values. The released model uses a value of 2.

Thank you very much for your answer!

[gobigrassland]

Hi,

This weight is used to balance the value of the two losses during the training process, so you can adjust it according to the actual values. The released model uses a value of 2.

github项目页图片，$\lambda$ 是和latent部分损失项相乘的。而train_codes代码部分则是反过来。是不是写错啦？

# Mask the top half of the image and calculate the loss only for the lower half of the image.
 image_pred_img = image_pred_img[:, :, image_pred_img.shape[2]//2:, :]
image = image[:, :, image.shape[2]//2:, :]    
loss_lip = F.l1_loss(image_pred_img.float(), image.float(), reduction="mean") # the loss of the decoded images
loss_latents = F.l1_loss(image_pred.float(), latents.float(), reduction="mean") # the loss of the latents
loss = 2.0*loss_lip + loss_latents # add some weight to balance the loss

[chunyu-li]

Hi,
This weight is used to balance the value of the two losses during the training process, so you can adjust it according to the actual values. The released model uses a value of 2.

github项目页图片，λ 是和latent部分损失项相乘的。而train_codes代码部分则是反过来。是不是写错啦？

# Mask the top half of the image and calculate the loss only for the lower half of the image.
 image_pred_img = image_pred_img[:, :, image_pred_img.shape[2]//2:, :]
image = image[:, :, image.shape[2]//2:, :]    
loss_lip = F.l1_loss(image_pred_img.float(), image.float(), reduction="mean") # the loss of the decoded images
loss_latents = F.l1_loss(image_pred.float(), latents.float(), reduction="mean") # the loss of the latents
loss = 2.0*loss_lip + loss_latents # add some weight to balance the loss

确实，我也感觉这里是作者写错了

[czk32611]

Hi,
This weight is used to balance the value of the two losses during the training process, so you can adjust it according to the actual values. The released model uses a value of 2.

github项目页图片，$\lambda$ 是和latent部分损失项相乘的。而train_codes代码部分则是反过来。是不是写错啦？

# Mask the top half of the image and calculate the loss only for the lower half of the image.
 image_pred_img = image_pred_img[:, :, image_pred_img.shape[2]//2:, :]
image = image[:, :, image.shape[2]//2:, :]    
loss_lip = F.l1_loss(image_pred_img.float(), image.float(), reduction="mean") # the loss of the decoded images
loss_latents = F.l1_loss(image_pred.float(), latents.float(), reduction="mean") # the loss of the latents
loss = 2.0*loss_lip + loss_latents # add some weight to balance the loss

我确认了一下代码，图画错了，代码没错。。。pixel维度的loss权重是2