In almost every intelligent tutoring systems (ITS) the main focus is to improve student's learning efficiency without considering the student engagement. Student's experience with the platform is important factor, lack of which leads to student dropout from sessions. While there have been conducted several researched about predicting student dropout from school or MOOC, predicting student dropout from study sessions is completely new area. In this paper specifically mobile environment was considered due to the fact that there are various factors that can cause student dropout from the session like applications, messaging and so on. Previous models for the student dropout have failed when predicting the student study session dropout in the mobile environment which is pretty self-explanatory. Study sessions are defined as the sequence of interactions for which the time difference is less than or equal to 1 hour. If this is not the case, then there would be study session dropout. The authors propose the transformer architecture for this task, namely, DAS which similar to transformers includes stacked encoders and decoders, they use SANTA dataset.