Comment by stingrae

the model is a set of weights, you can take a snapshot and test it. Reinforcement learning itself is largely testing and tuning.