Is there a way to go from the continuous variable quantum computer model to the qubit model? There are some papers such as  https://arxiv.org/abs/quant-ph/0008040 and https://arxiv.org/abs/1907.09832 that discuss going from qubit to the continuous model. But I'm curious if there is a way to go from the continuous variable model to the qubit model?
 A: There is a way because a set of $n$ qubits allows universal computation on Hilbert spaces of up to $2^n$ dimensions. But the word "universal" has hidden in it some issues surrounding computational complexity. The question becomes, can qubits efficiently simulate continuous variables. The answer to that is that if a continuous variable calculation can be approximated sufficiently well in some finite number of dimensions, then the simulation via qubits is efficient (i.e. no more complex than whatever algorithm was being invoked on the continuous system).
To be honest, I am only sure of that up to some finite degree; I am not quite certain and may be mistaken. Perhaps others will answer. But my high degree of confidence is based on the fact that simulation of continuous things using discrete things is part of the basic structure of quantum computing and has been proved to be efficient in a very wide sense. This is not to say that all quantum computations are polynomial; they are not. Grover's algorithm certainly is not, and NP complete problems are thought to be not. But qubits are as good at computing as anything else; that is the basic result. 
To go from a continuous model to a discrete one, you can simply adopt standard ideas in computer science. For example, take the lowest $p$ levels of each harmonic oscillator, or some finite set of coherent states, or something like that. Then write down all your operators in terms of matrices for your chosen basis. Then simulate that basis with qubits and translate the matrices to sequences of logic gates.
