Training with age from scratch

[ntkarim]

Hello thank you for the model you provided. I'm currently working on the age model with two classes and I trained adience as well on two classes and got 83.45% training accuracy. Then I'm trying to finetune that one with a dataset of mine and the training loss isn't going down and it stays static for all of 50k iteration. I kept the hyper parameters of the model the same as you have given. Should I try decreasing the bias_filler of conv2?

[GilLevi]

Thank you for your interest in our work.

When finetuning on your dataset, what was the learning rate that you used?

Best, Gil

[ntkarim]

Hello. I was using 0.001 at first same as yours. But then decided to change it to 0.0001. It worked although the network is getting overfitted.

[GilLevi]

Does the validation loss decrease?

[ntkarim]

its decreasing but validation loss is increasing after a certain point. so Im guessing its because of the model being overfitted?

[GilLevi]

Exactly. You should stop training when the validation loss is at the minimum.