My test loss increase after 10 epochs but the train loss converged well

Hi I used the resnet50 and I found that after 10 epochs, my test loss increased(before that it decreased) but the train loss converged. Can anybody offer me some solutions to solve this problem?

Your network is overfitting. I believe, your dataset for training is small. If yes, try to use larger dataset.

I don’t have bigger dataset. Are there some other solutions?

In this case, try to use data augmentation, and use learning rate decay. However, I recommend you to share more details about your cureent setups (e.g., task, model, dataset properties, lr rate, etc). Its will be helpful to understand the problem and debug it.