The title pretty much gets the whole question across. In particular, did Einstein work with the modern definition of smooth manifolds, or was that formalized later? Did he know about/use the coordinate-free approach, or did he only consider vectors/tensors to be "things that transform as <rule>"?...