I'm planning an RL project and I have to decide which RL framework do I use if any at all. The project has a highly custom environment, and testing different algorithms will be required to obtain optimal results. Furthermore, it will use a custom neural network, not implemented in the popular TensorFlow/PyTorch ML frameworks. Therefore, the framework should allow for customization with regard to approximation function (1) and the environment (2). The problem is that to my current knowledge, most of the framework allows only to work with a built-in environment. Does anybody know a framework that meets the two conditions (1) and (2)? Or anybody knows a review that contains information about framework in the context of those conditions?