DeepXDE is a library for scientific machine learning and physics-informed learning. It implements physics-informed neural networks (PINNs) and DeepONet architectures to solve forward and inverse problems involving ordinary and partial differential equations, fractional PDEs, and stochastic PDEs. The library supports multiple deep learning backends including JAX, PyTorch, TensorFlow, and PaddlePaddle, leveraging automatic differentiation to enforce physical constraints in the neural network loss functions.