how do I know if my model is overfitting from a learning gragh?

I assume orange is your validation loss. To have a good balance your validation loss and train loss should move together. Meaning, their loss values needs to be almost same. Visually, in the graph, there shouldn't be a huge gap between these lines. If your validation loss is much bigger than your train loss then you are encountering overfitting. If your train loss is much bigger than your validation loss then you have a underfitting issue. In the epoch 4, your train loss and validation loss seems okay (for this training visual at least). But after the epoch 2 (except in the epoch 4), you are generally facing overfitting. It would also be good to compare your train accuracy vs validation accuracy to better understand your training process. You can read this article for more.