Convert CellBox Tensorflow code to Pytorch

CellBox is a model for predicting cell states under drug perturbations, whose code base was originally written in Tensorflow 1. Since Tensorflow 1 has been replaced by Tensorflow 2 and Pytorch, the aim of the project is to recreate CellBox in Pytorch. There are three main components of the model: the data processing pipeline, the model architecture, and the training process, all of which will be sequentially converted into Pytorch. After the conversion is complete, the new Pytorch model is trained using the same training configurations as the Tensorflow model. The following deliverables define the proposal’s success: the trained Pytorch model should reproduce the same results as the current Tensorflow model; the new Pytorch code base should be configured to be able to run on multiple platforms (Linux, Mac, …); and the github should have clear documentations for installing and running the model.