One of the common ways in physics to predict properties of materials is to make a statistical analysis of many atomic configurations and forces. The forces could be predicted by solving a quantum mechanical differential equation with a huge amount of variables. Typically one would need millions of such calculations to predict such properties as diffusion constant, melting point, or battery performance. In our work, we suggested a direct and local DL model for atomic forces. We analyze the model's performance as a function of the input and show that one can ascertain physical attributes of the system from the analysis of the DL model's behavior.