I am training a segmentation model. I am training a UNet like architecture with a pre-trained encoder. I am using segmentation_models_pytorch for a model architecture. I am using 1000 images. My optimizer is Adam with a learning rate of 1e-3 and batch size 16.
However, my loss does not decrease beyond a point. I also tried using lr_scheduler. ReduceLROnPlateau but no luck. I trained the model for 1800 epochs but the loss did not decrease but ReduceLROnPlateau decreased the learning rate.