Causal representation learning is an emerging research area in machine learning. Dedicated to learning causal graphs from unstructured data. Usually this is done through either extracting high-level features through deep neural networks or other kinds of featurizers and learning causal graphical model based on the extracted features, or through an end-to-end network which involves extracting identifiable high-level features through some model such as ICA(independent component analysis) and using the learned representation in causal discovery directly. This project will be used for training and using deep learning models in my research in causal representation learning. In particular, currently we will be implementing sequential models like LSTMs and transformers to compare the performance to simpler but identifiable representations of models like multilayer perceptrons (MLPs).