Loss function during training

[yld1028]

Hello,

I hope you’re doing well. I have a question regarding the training process described in your article. I followed the instructions and started training the model, but I noticed that the loss function initially becomes negative instead of positive. I’m not sure if this is expected or if there might be an error in my implementation.

I have carefully reviewed the documentation and double-checked my code, but I haven’t been able to find a solution. Could you please help me understand if a negative loss value at the beginning of the training process is correct or not?

Thank you so much for your time and expertise.

[Tombs98]

Hello,

I hope you’re doing well. I have a question regarding the training process described in your article. I followed the instructions and started training the model, but I noticed that the loss function initially becomes negative instead of positive. I’m not sure if this is expected or if there might be an error in my implementation.

I have carefully reviewed the documentation and double-checked my code, but I haven’t been able to find a solution. Could you please help me understand if a negative loss value at the beginning of the training process is correct or not?

Thank you so much for your time and expertise.

In this work, we use PSNR loss following NAFNet. It will get smaller and smaller. You can see megvii-research/NAFNet#48 (comment) for more informations