update readme

README.md

@@ -173,6 +173,8 @@ param_groups = [
 # that has two components (base and classifier). Modify the code to adapt to your model.
 optimizer = torch.optim.Adam(param_groups, lr=args.lr, weight_decay=args.weight_decay)
 ```
+Of course, you can pass `model.classifier.parameters()` to optimizer if you only need to train the classifier (in this case, setting the `requires_grad` wrt the base model to false will be more efficient).
+
 
 ## References
 [1] [He et al. Deep Residual Learning for Image Recognition. CVPR 2016.](https://arxiv.org/abs/1512.03385)<br />