We report two methods for solving FBSDEs of path dependent types of high dimensions. Specifically, we propose a deep learning framework for solving such problems using path signatures as underlying features. Our two methods (forward/backward) demonstrate comparable/better accuracy and efficiency compared to the state of the art techniques. More importantly, we are able to solve the problem of high dimension which is a limitation in the conventional methods. We also provide convergence proof for both methods with the proof of the backward methods in the Markovian case.