Text this: Practical sparse data-driven constitutive modeling via transfer learning in physics-encoded neural networks.